Raman Spectroscopy for the Analysis of Thin CuInS Films€¦ · damentals of Raman spectroscopy...

Raman Spectroscopy for the

Analysis of Thin CuInS2 Films

von

Dipl.-Phys. Thomas Riedle

aus Wendlingen

Der Fakultat II (Mathematik und Naturwissenschaften)

der Technischen Universitat Berlin

zur Verleihung des akademischen Grades

D o k t o r d e r N a t u r w i s s e n s c h a f t

vorgelegte Dissertation

Berlin 2002

D 83

Arbeit eingereicht am: 14. Marz 2002

Promotionsausschuß:

Vorsitzender: Prof. Dr. Erwin Sedlmayr

Berichter: Prof. Dr. W. Richter

Prof. Dr. M. Ch. Lux-Steiner (FU-Berlin)

Prof. Dr. Ch. Thomsen

Tag der mundlichen Prufung: 14. Mai 2002

II

Kurzreferat

Dunne CuInS2 Filme werden als Absorbermaterial fur Solarzellen verwendet. Im

Rahmen dieser Arbeit wurden die Filme innerhalb weniger Minuten durch reaktives

Anlassen von Cu-In Metall-Vorlauferschichten in H2S-Gas hergestellt. Zum ersten Mal

wurde dafur ein schneller thermischer Prozess verwendet. Raman-Spektroskopie wurde

fur die Analyse der CuInS2 Filme eingesetzt.

Die phononischen Eigenschaften der CuInS2 Filme sind abhangig vom [Cu]/[In]

Verhaltnis, das fur die Reaktion angeboten wird. Fur [Cu]/[In] < 1 zeigen sich neben

den bekannten Phononen des Chalkopyrit-Gitters zusatzliche Moden bei 305 cm−1 und

60 cm−1. Diese Phononen werden auch fur Filme beobachtet, die durch reaktives An-

lassen kupferreicher Vorlauferschichten ([Cu]/[In] > 1) in H2S im Temperaturbereich

375 C - 475 C entstehen. Durch polarisationsabhangige Raman Messungen konnte

gezeigt werden, dass die Mode bei 305 cm−1 von einer symmetrischen Schwingung (A1)

stammt. Fur die zusatzlichen Moden ist eine polymorphe Struktur des Chalkopyritgit-

ters, die sogenannte CuAu-Ordnung, verantwortlich. Dies konnte anhand gruppentheo-

retischer Betrachtungen und Berechnung der Phononenfrequenzen fur das CuAu-Gitter

belegt werden.

Mittels kombinierter Raman- und Rontgenbeugungsanalysen konnte die Reaktions-

kinetik untersucht werden. Chalkopyrit- und CuAu-geordnete Strukturen entstehen

aus der Vorlauferschicht bereits in der fruhen Aufheizphase, wobei die CuAu-Ordnung

bevorzugt wird. Der Anteil an CuAu-Ordnung nimmt bei weiterer Temperatur-

erhohung ab und verschwindet schließlich. Das uberschussige Kupfer bildet Cu9S5

und CuS an der Oberflache. Diese binaren Cu-S Phasen konnen durch chemisches

Atzen restlos entfernt werden.

Solarzellen auf Basis der hergestellten CuInS2 Schichten wurden anhand von Strom-

Spannungs-Kennlinien in Abhangigkeit der Herstellungsparameter charakterisiert. Es

konnte gezeigt werden, dass Konversionsverluste in den Solarzellen mit dem Auftreten

von CuAu-geordneten Domanen verbunden sind. Durch reaktives Anlassen von se-

quentiell aufgedampften Cu-In Schichten in H2S bei 525 C fur 5 Minuten konnten

einphasige CuInS2-Chalkopyritfilme hoher kristalliner Qualitat prapariert werden. Auf

Basis dieser Filme wurden Solarzellen mit bis zu 11 % Wirkungsgrad hergestellt. Der

verwendete Aufbau fur die Raman-Messungen zusammen mit den Daten dieser Arbeit

kann fur die ex-situ Qualitatskontrolle von CuInS2-Absorber Filme genutzt werden.

Daruber hinaus sind die Grundlagen fur die Entwicklung eines Raman-Aufbaus fur die

in-situ Kontrolle des CuInS2-Wachstums gelegt.

III

Eidesstattliche Erklarung

Hiermit erklare ich an Eides Statt, daß ich bei der Anfertigung dieser Arbeit keine

anderen als die angegebenen Hilfsmittel benutzt habe. Die Dissertation ist bis auf die

gekennzeichneten Teile noch nicht veroffentlicht worden.

Ich habe weder fruher noch gleichzeitig ein Promotionsverfahren bei einem anderen

Fachbereich bew. einer anderen Hochschule beantragt.

Thomas Riedle

Berlin, den 14. Marz 2002

IV

To Iva and Kaja

V

Contents

1 Material Properties 3

1.1 Structural Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.1.1 Polymorphous Structures . . . . . . . . . . . . . . . . . . . . . . 5

1.1.2 Brillouin-Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.2 Electronic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.2.1 Crystal field splitting . . . . . . . . . . . . . . . . . . . . . . . . 10

1.2.2 Optical Properties . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.2.3 Exciton properties . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.3 Vibrational properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.4 Phase Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.4.1 Cu-In System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.4.2 Cu2S-In2S3 Phases . . . . . . . . . . . . . . . . . . . . . . . . . 18

2 Experimental procedures 20

2.1 Solar Cells Based on CuInS2 . . . . . . . . . . . . . . . . . . . . . . . . 21

2.1.1 Absorber Preparation . . . . . . . . . . . . . . . . . . . . . . . . 22

2.2 Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.2.1 Macroscopic Theory of Inelastic Light Scattering by Phonons . . 24

2.2.2 Raman Tensor and Selection Rules . . . . . . . . . . . . . . . . 25

2.2.3 Microscopic Theory of Raman Scattering . . . . . . . . . . . . . 26

2.2.4 Experimental Setup for Raman Scattering . . . . . . . . . . . . 27

3 Raman Spectroscopy of Thin CuInS2-Films 29

3.1 Laser Induced Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.2 Lateral Inhomogeneities of Polycrystalline CuInS2 Films . . . . . . . . 31

3.3 Resonant Raman Scattering . . . . . . . . . . . . . . . . . . . . . . . . 33

3.3.1 Raman Scattering at Excitation Energies above the Fundamental

Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4 Non-Chalcopyrite Phonon Modes in CuInS2 37

4.1 Observation of CuAu Order and Theoretical Considerations . . . . . . 38

VII

VIII

4.1.1 Observation of Non-Chalcopyrite Phonon Modes . . . . . . . . . 38

4.1.2 Raman Selection Rules . . . . . . . . . . . . . . . . . . . . . . . 39

4.1.3 Group Theoretical Analysis . . . . . . . . . . . . . . . . . . . . 40

4.1.4 Lattice Dynamics of the CuInS2 - CuAu Structure . . . . . . . . 44

4.2 Dependence on Stoichiometry . . . . . . . . . . . . . . . . . . . . . . . 46

5 Reactive Annealing in H2S 51

5.1 Dependence on Sulfurization Temperature: Phonon Modes for the Ab-

sorber Front Side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.2 Dependence on Sulfurization Temperature: Phonon Modes for the Ab-

sorber Back Side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

5.3 Surface Segregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

5.4 Phase Formation by H2S . . . . . . . . . . . . . . . . . . . . . . . . . . 61

5.5 Photovoltaic Performance: Dependence on Morphology and Structure . 62

6 Summary 67

A CuS Phase Diagram 68

Introduction

There is a strong demand for renewable energies due to the limited availability of fossil

and nuclear fuels and due to growing environmental problems. Photovoltaic energy

conversion has the potential to contribute significantly to the electrical energy gen-

eration in the future [1]. Currently, the cost for photovoltaic systems is one of the

main obstacles preventing production and application on a large scale. A substantial

decrease in production costs for modules, and therefore in overall system cost, is ex-

pected from the development of thin film solar cells. This is the background for the

strong research interest in materials suitable for thin film solar cells like amorphous

silicon, CdTe and Cu(In,Ga)Se2 [2].

Conversion efficiencies of solar cells based on I-III-VI2 chalcopyrite compounds have

been substantially improved over the last years. Polycrystalline solar cells with a

Cu(In,Ga)Se2 absorber reached recently up to 18.8 % [3].

CuInS2 is a Se-free compound from the chalcopyrite family suitable for solar cells due

to its high optical absorption and the direct band gap at 1.5 eV. In principle, solar cells

based on sulfur chalcopyrites like CuInS2 have the same potential for high efficiencies

as those based on selenium chalcopyrites. The development of CuInS2 is attractive,

because the problematic selenium is substituted by the non-toxic sulfur. The open cir-

cuit voltage of CuInS2 solar cells can theoretically be higher (1.2 V) than the voltage of

Cu(In,Ga)Se2 cells. At the same time photo current is lower. This is advantageous for

the serial connection of multiple cells in a module. CuInS2 films can be fabricated in

a fast and robust process granting high throughput in an industrial process. However,

the efficiency of CuInS2 solar cells is up to now limited by the open circuit voltage

which is far below the theoretical value. The best reported conversion efficiency for

polycrystalline CuInS2 cells is 12.7 % up to now [4, 5].

From a technological point of view, fast processes are desirable for high production

output. Rapid thermal processing (RTP) systems equipped with halogen lamp arrays

were successfully applied for the synthesis of CuInSe2 films [6, 7]. Annealing of Cu-

In precursors in a H2S RTP system has been established in this work. The provided

tight control of process parameters was exploited for systematic variations. Vibrational

properties of the absorber structure were studied in dependence on the process para-

meters.

1

2

A lot of research has been carried out on chalcopyrite materials since the early seven-

ties and there are numerous reports about their vibrational properties [8]. However,

information on CuInS2 is limited. Most of the authors [9, 10, 11] studied CuInS2 single

crystal prepared under thermodynamical equilibrium conditions for the determination

of fundamental vibrations. Just a few authors [12, 13] reported on polycrystalline thin

films for solar cell applications which are usually prepared under non-equilibrium con-

ditions. This work focuses on the vibrational properties of of such films.

This thesis is organized as follows: In Chapter 1, the known structural, electronic and

vibrational properties of CuInS2 were compiled from literature data.

In Chapter 2, experimental procedures are described. The design of a CuInS2 solar

cell, preparation procedures and a description of the RTP system will be given. Fun-

damentals of Raman spectroscopy necessary for the interpretation of the experimental

data will be introduced followed by a description of the used setup. Most spectra in

this work were recorded by means of the micro-Raman technique for lateral resolution

and higher laser power density.

The laser light irradiated onto the sample for Raman spectroscopy may alter chemical

or structural properties. The upper limit of the laser power density for non-destructive

measurements has been determined and will be given in Chapter 3. The consequences

of the polycrystalline character of the films on Raman shift and peak width will be

discussed on basis of a Raman map.

Enhancement in Raman scattering intensities can be observed by tuning the incident

laser to resonate with a electronic transition. This is known as resonant Raman scat-

tering. A narrow resonance curve for CuInS2 films at excitation energies close to the

band gap will be presented. It will be shown that the incident photons are in resonance

with bound excitons close to the band edge.

A report about the simultaneous observation of chalcopyrite and non-chalcopyrite

phonon modes will be given in Chapter 4. The non-chalcopyrite mode at 305 cm−1 is

A1-symmetric according to the polarisation depend Raman measurements. The non-

chalcopyrite phonon modes can be attributed to CuAu-ordered CuInS2. It will be

shown that results from group theoretical analysis and phonon frequency calculations

performed in this work are in agreement with the presented experimental results.

CuInS2 phase formation by reactive annealing of Cu-rich precursors in H2S gas is the

subject of Chapter 5. Intermediate phases and surface segregations were identified by a

combination of Raman and X-ray diffraction measurements. The dependence of solar

cell efficiencies on the morphology and the crystal structure of CuInS2 films will be

discussed. It will be demonstrated how high quality single phase CuInS2 chalcopyrite

films can be derived from evaporated Cu-rich precursor films by choosing optimized

H2S annealing conditions.

Chapter 1

Material Properties

For this thesis the structural properties of thin CuInS2 layers were analyzed by Raman

spectroscopy. In the course of the work the existence of a new phase of CuInS2 was

proven. In this chapter, an introduction to structural and electronic properties of this

material will be provided as a basis for the interpretation of the experimentell results.

CuInS2 belongs to the family of I-III-VI2 chalcopyrite compounds. The members of

this family are related to each other by their chalcopyrite crystal structure. A variety of

different electronic properties, i.e. band gaps, result from the elements which build the

compound. Throughout research history on this class of materials a lot of fundamental

insights were derived by comparative studies [14]. Comparing the band gaps of Cu-

III-VI2 chalcopyrites with other I-III-VI2 compounds gave valuable hints about the

electronic contributions of Cu 3d orbitals to the valence band of Cu-III-VI2 materials.

The structural properties of CuInS2 will be introduced in Section 1.1 by discussing the

chalcopyrite structure. This will be followed by stability studies performed by Wei et

al. [15] predicting the occurence of a CuAu ordered phase in CuInS2. Subsequently

the Brillouin-zone of the chalcopyrite lattice will be presented for the discussion of the

electronic properties.

The electronic properties of CuInS2 will be discussed in this chapter, not only in view

of its photovoltaic application. The Raman scattering process is indirect, involving

virtual or real electronic transitions. The results from resonant Raman measurements

in this work were explained by taking the related electronic transitions into account.

The relation between the structural properties of chalcopyrites and the band gap will

be considered in Section 1.2. The effect of the crystal electric field on chalcopyrites and

especially on CuInS2 will be outlined thereafter. The band structure of CuInS2 and

possible band-band transitions will be discussed on the given basis. Subsequently, the

luminescence spectrum of CuInS2 single crystals will be reviewed. Special emphasis

was given on the excitonic properties of the material as an exciton was involved in

resonant Raman scattering at energies close to the band gap.

Finally, known vibrational properties of CuInS2 were collected from the literature and

3

4 Chapter 1. Material Properties

will be provided in Section 1.3. More details will be added throughout this thesis.

1.1 Structural Properties

There are 36 known ternary AIBIIICV I2 chalcopyrite semiconductors where A=Cu,Ag,

B=Al,Ga,In,Ti and C=S,Se,Te. The chalcopyrite structure can be systematically con-

structed starting from a cubic face centered structure. By arranging two units along

a diagonal line trough the cubes and shifting them, in terms of the basis vectors by

(a/4,a/4,a/4) the diamond structure is obtained. The zinc-blende structure can be de-

rived by occupying the (001) planes in the diamond structure with two different sorts

of atoms as depicted in Figure 1.1a.

(b) Chalcopyrite

A B CE F

(a) Zinc-blende

x

y

z

ab

c

Figure 1.1: (a) Zinc-blende unit cell, space group T2d. (b) Chalcopyrite unit cell, space

group D122d.

Finally, the chalcopyrite structure can be obtained by doubling the zinc-blende

structure along the z-axis and filling the lattice sites according to the following: The

anions remain at their stites and every second (001) plane is occupied by cations as

shown in Figure 1.1b. In consequence, each C anion is coordinated by two A and two

B cations and each cation is tetrahedrally coordinated by four anions.

The observed structural features from real chalcopyrite compounds are slightly

different from those obtained theoretically from this construction rules. The unique

properties of the chalcopyrites are related to three differences with respect to the

zinc-blende structure: First, there are two cation sublattices rather than one, leading

1.1. Structural Properties 5

to the existence of two basic chemical bonds A-C and B-C, with generally unequal

bond lengths RAC 6= RBC . Second, the unit cell is tetragonally distorted with a

distortion parameter η = c/2a 6= 1. Third, the anions are displaced from the ideal

tetrahedral site u0 = 1/4 by an amount u in direction of the x-axis. The structural and

electronic properties of the chalcopyrites are governed by the added structural (η, u)

and chemical (A 6= B) degrees of freedom relative to their binary analogs [16]. Struc-

tural and optical properties of selected chalcopyrite materials are compiled in Table 1.3.

1.1.1 Polymorphous Structures

Many solids with the same composition can appear in different crystal structures un-

der different thermodynamical conditions. This phenomenon is referred to as poly-

morphism [17]. A set of polytypes of the chalcopyrite structure was theoretically

constructed such that the electron counting rule is obeyed. Formation energies and

band structure of CuInSe2 and CuInS2 polytypes were determined by first-principles

calculations by Wei et al. [18, 15]. It was shown that the CuAu-like ordered structure

is the most possible to occur. It is referred to as the CuAu-like structure in analogy to

the structure of CuAu mixed crystals [17]. An illustration is given in Figure 1.2.

The anion sublattice is conserved in the CuAu structure and the cation order is

A B C

Chalcopyrite CuAu

ab

c

Figure 1.2: The chalcopyrite unit cell and the polymorphous CuAu-structure [15].

changed such that the A2B2 coordination is conserved. The unit cell of the CuAu-


ordered structure is given in Figure 1.3. The lattice type is primitiv tetragonal and

Figure 1.3: Unit cell of the CuAu-like chalcopyrite polytype.

the corresponding space group is P 4m2 [19]. The Wyckoff positions are given in Table

1.1. There is one type A-atom, one type B-atom and two type C-atoms in the unit

cell. The corresponding sites are c,a and g.

Table 1.1: Multiplicity, Wyckoff letter, site symmetry and positions of the atoms in

the CuAu-like structure [20].

Multiplicity Wyckoff Symmetry Coordinates

letter

2 g 2mm 0, 12, 1

412, 0,−1

4

1 c 4m2 12, 1

2, 1

2

1 a 4m2 0,0,0

An exceeding small formation energy difference ∆Eform = −1.95 meV/atom was

found between chalcopyrite and CuAu like phases of CuInS2. Similar results were

found for CuInSe2 where ∆Eform = −2.05 meV/atom. The coexistence of CuAu-like

phases in nominally chalcopyrite CuInS2 and CuInSe2 was predicted. In contrast, for

CuGaSe2 ∆Eform = −9.05 meV/atom was found. The CuAu-like phase is therefore

less likely for CuGaSe2.

Band gap energies are affected by the transition from chalcopyrite (CH) to poly-

morphous structures. Calculations resulted in EG(CH) − EG(CuAu) = 30 meV for

CuInS2 and in EG(CH) − EG(CuAu) = 46 meV for CuInSe2 [15]. The small dif-

ferences suggest that formation of polytypes in this compounds has little effect on

1.1. Structural Properties 7

their electrical and optical properties. The situation is different for CuGaSe2 were

EG(CH) − EG(CuAu) = 232 meV was found. A larger effect was expected for this

compound.

1.1.2 Brillouin-Zone

The electronic band structure of semiconductors is given in k-space. The Brillouin

zone of chalcopyrites is presented here for the later discussion of the band structure.

The Brillouin zone and its relationship to that of the zinc-blende is given in Figure 1.4.

The corresponding primitive cell contains eight atoms (2·I-III-VI2) instead two found

in zinc blende. Consequently the Brillouin zone reduces its volume by a factor 4. Sets

of four different wavevectors of the original zinc-blende Brillouin zone fold into a single

point of the four times smaller chalcopyrite Brillouin zone. The three main symmetry

points and their origins in zinc-blende are summerized in Table 1.2.

TD

G

TD

kZ

GX

NL

TX

kX

GW

NS

TX kY

GW

Figure 1.4: Brillouin zone of chalcopyrite (CH) and its relationship to that of zinc-

blende (ZB). The dotted polyhedra show the ZB reciprocal-space regions that

fold into the CH Brillouin zone. Symmetry points are labeled AB, where A

and B refer to the CH and ZB symmetries, respectively [21].


Table 1.2: Symmetry points of the chalcopyrite (CH) Brillouin zone and their origins

in the zinc-blende (ZB) zone [21].

CH-symmetry point Related symmetry point in zinc-blende

Γ(000) Γ(000) X(002) W(201) W(021)

T(001) ∆(001) ∆(001) X(200) X(020)

N(110) L(111) L(111)∑

(110)∑

(110)

1.2 Electronic Properties

The band-gap of I-III-VI2 chalcopyrites is controlled by two factors. First, there is a

pure structural factor due to the existence of a displacement from the ideal tetrahedral

site u0 = 1/4. This parameter controls the band gap in the system. Even a small

increase in u from its ideal zinc-blende value leads to a substantial ionic polarization of

the bonds and consequently to a dramatic increase in the band gap [22]. This can be

verified by inspection of Table 1.3 where u is listed together with the band gap energies

of some Cu-III-VI2 compounds.

Table 1.3: Values of the cubic lattice constants a and c, the tetragonal distortion

parameter η=c/2a, the anion displacement parameter u and the observed

lowest band gaps at T = 300 K [16], [22].

Ternary a=b c η u EG

(A) (A) (eV)

CuAlS2 5.334 10.444 0.979 0.275 3.49

CuGaS2 5.356 10.433 0.974 0.275 2.43

CuInS2 5.523 11.118 1.0065 0.214 1.53

CuAlSe2 5.602 10.946 0.977 0.269 2.71

CuGaSe2 5.614 11.032 0.9825 0.250 1.68

CuInSe2 5.784 11.614 1.004 0.224 1.04

The second factor is a electronic one. For the Cu-III-VI2 compounds a great influ-

ence of the novel atom 3d states on the valence band was found. These states hybridize

with the p states of the group VI elements. As the d states are found in the upper half

of the valence band they are partly responsible for the reduction of the band gap.

A schematic band structure of CuInS2 and the contributions of the atomic orbitals is

given in Figure 1.5.

1.2. Electronic Properties 9

~17 eV

~12 eV

~7 eV

~5 eV

0 eV

-1,5 eV

-EB.v

1

2

3

Conduction Band

Optical Gap

Upper Valence Band

In 4d

S 4s

In 4p

S 3p

Cu 3d

S 3p

S 3p

In 5s

T NG

Figure 1.5: Schematic band structure of CuInS2. The contributions of the atomic en-

ergy levels are indicated on the right. Shades areas denote the major sub-

bands, and boxed numbers mark the three internal gaps [16].

The valence band is separated in two parts. There is the upper valence band

reaching 5 eV and a lower part at 7 eV. Cu 3d and S 3p orbitals from the Cu-S bond

contribute to the upper valence band whereas S 3p and In 4p from the In-S bond form

the lower valence band. At around 12 eV a band is built from the S 4s states and

a small band is set up by the In 4d orbits. S 3p and In 5s orbits contribute to the

conduction band [16]. A more detailed band structure will be discussed in Section

1.2.2.


1.2.1 Crystal field splitting

The symmetry of an atom in free space will be reduced when it is placed in a crystalline

environment. The potential of the crystal causes a lifting of the degeneracies of the

atomic energy levels which become split by the crystal electric field. The crystal field

acts on the orbits of the electrons and will split the degeneracy of the free atom.

The influence of the tetrahedral field on the valence band energies of I-III-VI2 com-

pounds can be explained in a simple model given in Figure 1.6. The degenerate energy

Free

atoms

Spin-

orbit

Tetrahedral

field

cE

Ev

1

+

s6

8

7

7

8

8

15

15

12

p

d

G

G

G

G

G

GG

G

G

G

Figure 1.6: Scheme of the expected energy levels of valence band states in a tetrahedral

field [23].

eigen-values of the orbital states are drawn on the left hand side. The model does

not take into account the intermixing of p and d states in the valence band. Thus the

effect of the crystal field given in this scheme is legal only for the separated orbital

wave functions. However, the principles can be discussed within this sketch. The up-

permost p-levels of the valence band are depicted. In a tetrahedral field and due to

the spin-orbit coupling the p-levels will split in two levels (Γ8, Γ7) and the d-levels into

three levels (Γ7, Γ8, Γ8). The intermixing of p and d orbits causes two effects. First,

the uppermost Γ15 levels will be raised to higher energies. In consequence the band-

gap will be reduced. Second, the spin-orbit splitting of the uppermost valence bands

will be reduced, because the negative spin-orbit parameter (Γ8 − Γ7 splitting) of the

d-levels partially cancels the positive spin-orbit parameter of p-levels. The correlation

of the effects was used to estimate the degree of p-d hybridization from the spin-orbit

splitting of the uppermost valence bands. For CuInS2 and CuGaS2 a contribution of

45 % and 36 % respectively, of d-like states was found [24].

From all members of the the I-III-S2 family, spin-orbit splitting was solely reported for

CuInS2. The splitting was -20 meV. Furthermore it is the only I-III-VI2 chalcopyrite

which was not subject to crystal field splitting.


1.2.2 Optical Properties

The optical properties of a semiconductor are closely linked to its electronic band

structure. Therefore the band structure of I-III-VI2 chalcopyrites will be discussed.

It was calculated for Cu-based ternary semiconductors within the density-functional

formalism [16] taking into account the p-d hybrids of the valence band and the struc-

tural peculiarities discussed in Section 1.1. A generic band structure for Cu-III-VI2chalcopyrites is given in Figure 1.7.

T G N

G3

G2

G1

G5

G5

T + T3 4

T5

T5

T + T1 2

E(X )G

G4

E( X)G

E'( X)G N1

N1

N1

Energ

y(e

V)

VB

CB

-2

0

2

4

Figure 1.7: Band structure of CuInS2. Dashed and solid arrows represent optical tran-

sitions allowed in E ‖ c and E ⊥ c, respectively [21].

Critical point parameters of optical transitions can be derived by analysis of the

dielectric function ε(ω). In general, the electronic and optical properties of semiconduc-

tors are not isotropic. Therefore the dielectric function is given as complex second-rank

tensor and its components must be determined for different polarizations of the incident

light [25]. The features observed in ε(ω) are usually correlated to interband transitions

at high symmetry points in the band structure. Such measurements and analysis were

performed by Alonso et al. [21]. The measured real and imaginary parts of ε(ω) are

reproduced in Figure 1.8.


3

4

5

6

7

8

9

<ε 2>

E c E || c

<ε 1>

Energy (eV)

1 2 3 4 5

1

2

3

4

5

6

7

Γ

∆Γ

E2

E( X)

E(X )

E1(A)

E( X)

E0

E c E || c

104

105

106

E c

E || c

Pe

ne

tra

tio

n d

ep

th L

(n

m)

Energy (eV)

Ab

so

rpti

on

a (

1/c

m)

1 2 3 4 5 610

100

E c

E || c

Figure 1.8: On the left: Real and imaginary parts of the complex dielectric function

of CuInS2 for normal and perpendicular laser light incidence [21]. On the

right: Absorption coefficients and penetration depth calculated from ε2.


The identified optical transitions, their energies and polarizations are compiled in

Table 1.4. The letters A and B in brackets denote the different energies being found

Table 1.4: Main optical transitions energies (in eV) and their polarizations above the

fundamental edge in CuInS2.aRef. [26] (77 K). b Ref. [21] (300 K).

Label E‖c E⊥c

E0(A) 1.530b 1.530b

E0(B) - 1.530b

E(ΓX) 2.75(8)b

3.099a 3.087a

E1(A) 3.27(1)b 3.27(5)b

3.427a 3.246a

E(XΓ) 3.6(1)b 3.5(1)b

3.655a 3.669a

E(∆X)

E1(B) 3.94(5)b 3.9(1)b

4.053a 4.091a

E’(ΓX) 4.4(1)b 4.4(2)b

E2(A) 4.8(1)b 4.7(1)b

5.038a

E2(B) 5.09(3)b 5.05(3)b

5.033a

due to crystal field splitting. The expected difference between E0(A) and E0(B) of

20 meV was not observed, but this was attributed to the limited resolving power of

the spectrograph. Two letters in brackets indicate the high symmetry point in the

Brillouin zone and its origin in the Brillouin zone of zinc-blende (compare with Figure

1.4). From symmetry considerations it is known that that the transition E0(B) is

forbidden in E ‖ c polarization, but allowed in E ⊥ c.

Information about the absorption coefficient α and the penetration depth L of light

in the material can be derived from the dielectric function. First, the refractive index

n and the extinction coefficient κ were calculated from the relation

n(ω) = n + iκ =√

ε1(ω) + iε2(ω) (1.1)

where n is the complex refractive index. The absorption coefficient is given by

α =4πκ

λ0

(1.2)


and the intensity of the light in a given depth x was calculated using the law of Lambert-

Beer [25]:

I(x) = I(0)e−αx. (1.3)

The depth in which 1/e of the initial light intensity I(0) remains is called the penetration

depth L and is given by 1/α. The penetration depth is plotted as a function of the

photon energy in Figure 1.8. The value is 100 nm for the green laser line (2.43 eV).

The absorption coefficient in the visible spectrum of the light is about 105 cm−1.

1.2.3 Exciton properties

Resonant Raman scattering from CuInS2 was performed in this work and the results

will be presented in Chapter 1. A narrow resonance curve for the A1-phonon mode

was found for energies close to the fundamental band gap. For the explanation of the

resonance behaviour the involvement of excitons was discussed. Luminescence spectra

of CuInS2 and the occurence of excitons therein were reviewed for this purpose.

Free excitons are electron-hole pairs which are weakly bound by the attractive Coulomb

interaction. They can be described in the hydrogen model [27]:

Eexz(n) =µ e4

8 ε2h2

1

n2(n = 1, 2, 3...) (1.4)

1

µ=

1

me

+1

mh

where me and mh are the reduced electron and hole mass, respectively.

The discrete energy states Eexz(n) for CuInS2 were calculated from literature values

[28, 29, 24], using me = 0.16 m0, mh = 1.3 m0, ε = 11ε0:

Eexz(n) = 19.51

n2meV (1.5)

In contrast, bound excitons are localized to charged centers and their binding energy

is higher than the binding energy of the free excitons [27].

The low-temperature photoluminescence spectrum of a CuInS2 single crystals is shown

in Figure 1.9.

1.3. Vibrational properties 15

~~~~

Re

lati

ve

Em

iss

ion

Inte

ns

ity

CuInS2

1.40 1.45 1.50 1.52 1.53 1.54

Photon Energy (eV)

Wavelength (nm)

900 880 860 840 820 810 540 520 500 500 495

2.30 2.40 2.48 2.50 2.52

~~~~

Re

lati

ve

Em

iss

ion

Inte

ns

ity

Photon Energy (eV)

Wavelength (nm)

CuGaS2

2 K 2 K

E=

1.5

35

eV

ex

z

E=

2.5

03

eV

ex

z

Figure 1.9: Photoluminescence spectrum of a CuInS2 single crystals recorded at 2 K

[30].

The sharp line at 1.535 eV was attributed to the decay of free excitons with

the hole belonging to the uppermost valence band. The band gap of CuInS2 at

2K was 1.555 eV. Thus the binding energy of the free exciton was Eexz = 20 meV.

This is in accordance with the value calculated from the hydrogen model in (1.5).

The other sharp lines in the spectra are excitons bound to impurities or defects.

A broad luminescence band was observed at the low energy side of the spectra. It

was attributed to donar-acceptor transitions as the peaks shift systematically with

increasing excitation intensities [31].

1.3 Vibrational properties

There are only a small number of publications on the vibrational properties of CuInS2.

The first report on some of the high frequency modes was given by Koschel [9] in 1975.

The same author performed further measurements and added low frequency data in

the same year [10]. All modes predicted by group theory (refer to Section 4.1.3) were

observed by Koschel, besides the three B1 modes. Until today there are no reports

about this three modes. This might be attributed to a weak electron-phonon coupling

expressed in small off-diagonal elements d in the Raman tensor for the B1-mode (refer

to Table 2.2.2). Data, consistent with those from Koschel, were provided by Bacewicz

[11]. An elaborated review of the literature data for the vibrational frequencies of 14


chalcogenides was recently published by Ohrendorf und Hausler [8].

Mode assignments performed in this work will thus refer to the frequencies compiled

in Table 1.5. As the bondings in CuInS2 are highly ionic LO-TO splitting due to the

Frohlich interaction [25] was observed for the B2 and E modes.

Table 1.5: Symmetry and vibrational frequencies of CuInS2 phonon modes [10].

Asterisks indicate modes which are measured at T = 78 K.

Symmetry Frequency Symmetry Frequency

(cm−1) (cm−1)

A1 294∗ E1TO 321

B11 n.o. E2

TO 295

B21 n.o. E3

TO 244

B31 n.o. E4

TO 140∗

B12 TO 323 E5

TO 88∗

B22 TO 234 E6

TO 67∗

B32 TO 79∗ E1

LO 339

B12 LO 352 E2

LO 314

B22 LO 266 E3

LO 260

B32 LO 79∗ E4

LO 140∗

E5LO 88∗

E6LO 67∗

1.4. Phase Equilibria 17

1.4 Phase Equilibria

Thin polycrystalline CuInS2 layers can be prepared by a variety of methods. The most

important are the physical coevaporation of the elements in a vacuum system [37] and

the two stage sequential process. In the latter, copper and indium are sequentially

deposited by sputtering or evaporation. In a second stage the metal stack is sulfurized

by conventional thermal processing (CTP) either in elemental sulfur [38] or in H2S [39].

In this work a new method of sulfurization was established, the reactive annealing by

rapid thermal processing (RTP) in H2S. The CuInS2 phase formation during the H2S-

RTP process will be discussed in Chapter 5.4 with respect to the experimentell results

of this work. For this purpose the ternary (Cu-In-S)-phase diagram will be considered

here. Thereafter, the Cu-In and Cu2S-In2S3 tie lines will be reviewed.

A schematic ternary (Cu-In-S)-phase diagram is given in Figure 1.10a. For the sake of

clarity only binary phases on the Cu2S-In2S3 and CuS-InS intersections are depicted.

Besides CuInS2 only one more ternary phase CuIn5S8 in spinel structure was observed

in the (Cu-In-S)-system [40].

Cu

S

Cu S2

CuS InS

In S2 3

CuInS2

CuIn S5 8

In

Figure 1.10: a) Phase diagram of the (Cu-In-S)-system. In the schematic representa-

tion only solid state phases on the Cu2S-In2S3 and CuS-InS intersection

are depicted.

1.4.1 Cu-In System

The Cu-In phase diagram is given in Figure 1.11. The stable phases at room

temperature are Cu, Cu7In3 (δ phase), Cu16In9 (η phase), CuIn2 and In. CuIn2 was

not drawn into the phase diagram, because of uncertainties about the stability range

of the alloy. But it is known that CuIn2 forms below room temperature and is stable

up to 148 C [41]. The melting point of In is 156 C while Cu11In9 is stable up to 307C. Cu16In9 (η phase) undergoes a phase transition to the η’ phase between 307 C


and 389 C.

Figure 1.11: Cu-In binary phase diagram [42].

The given phase diagram is valid for bulk materials prepared under thermal equi-

librium conditions. The Cu-In phase formation in thin films was studied by several

authors, but no additional phases were reported for this films [43, 44, 45].

At room temperature the alloy formation is governed by Cu diffusion into In, whereas

grain boundary diffusion of In into the Cu-layer is the dominant transport mechanism

above T = 150 C [46].

1.4.2 Cu2S-In2S3 Phases

The phase diagram of the binary system Cu2S-In2S3 is given in Figure 1.12. All com-

pounds occurring in this system are summarized in Table 1.6.

Two semiconducting phases CuInS2 and CuIn5S8 appear in the diagram. CuInS2

exists in three modifications, up to 980 C in the chalcopyrite structure, between 980C and 1045 C in the zincblende structure and above 1045 C up to the melting

point at 1090 C in a still unknown structure, possibly wurtzite [40]. The second

semiconductor, CuIn5S8 has the spinel structure over the whole temperature range of

20 C to the melting point at 1085 C.

1.4. Phase Equilibria 19

Figure 1.12: Phase diagram of the (Cu-In-S)-system along the Cu2S-In2S3 tie line

[40]. The single phase regions are indicated by their respective symbol.

The two phase regions, which lie in between the single phase regions are

not indicated.

Table 1.6: Compounds occurring in the system Cu2S-In2S3 with their different modifi-

cations and transition temperatures [40].

Compound Modification Transition

temperature (C)

Cu2S α1 tetragonal 104

α2 hexagonal 450

α cubic 1125 (m.p)

CuInS2 γ chalcopyrite 980

δ zincblende 1045

ξ wurtzite 1090 (m.p.)

CuIn5S8 ε spinel 1085 (m.p.)

In2S3 η1 defect-spinel superstructure 420

η2 defect-spinel-structure 755

η layered structure 1090 (m.p.)

Chapter 2

Experimental procedures

The experimental work performed for this thesis focused on preparation as well as on

structural analysis. Rapid thermal processing in H2S gas was established as a novel

method for sulfurization of Cu-In metal stacks in order to obtain high quality CuInS2

absorber films for efficient solar cells. The unique feature of this process is the well

controlled film growth. Exploiting this feature, vibrational properties of CuInS2 films

were analyzed by Raman spectroscopy in dependence on the H2S sulfurization param-

eters.

In this chapter the preparation techniques will be described. The structure of a het-

erojunction solar cell based on a thin CuInS2 absorber film will be presented, followed

by a brief summary of process steps used for cell fabrication. The preparation of the

absorber will be explained more detailed. The one step simultaneous evaporation pro-

cedure and the two step sequential process will be presented. The sequential process

requires an sulfurization step to transform the metallic precursors in CuInS2. An out-

line of the RTP system for reactive annealing in H2S and its features will be given.

Raman spectroscopy will be the subject of the second part of this chapter. Fundamen-

tals of inelastic light scattering will be considered from the macroscopic and microscopic

view. The experimental setup will be presented and its features will be discussed.

20

2.1. Solar Cells Based on CuInS2 21

2.1 Solar Cells Based on CuInS2

Thin film solar cells can be prepared by the use of the high absorbing CuInS2 chal-

copyrite. The typical device structure is shown in Figure 2.1.

Light

(Ni : Al) 1 m

(ZnO : Al) 500 nm

CdS 30 - 80 nm

2CuInS 3 m

Molybdenum 1 m

Soda lime 2 mmglass

p-type

n-type

Figure 2.1: Structure of a thin film

CuInS2 solar cell.

Conventional soda lime glass serves as device substrate. The molybdenum film

is used as back contact. The light is absorbed by the p-type CuInS2. Due to the

high absorption coefficient of CuInS2 sun light is completely absorbed within the three

microns. The heterojunction is completed by a thin n-type CdS buffer layer and a

n/n+-type ZnO bilayer. The band gaps of CdS and ZnO are 2.4 eV [47] and 3.2 eV

[48], respectively. The combination of both is referred to as the ”window” because it

is almost transparent for the sun light. A Ni/Al front grid is used as front contact for

laboratory small scale devices. An overview on processing steps for the preparation of

the above introduced CuInS2 solar cells is given here:

The substrate glass is chemically cleaned and dried in a hot air stream in order to re-

move contaminations from the surface [49]. The molybdenum back contact is deposited

by e-gun evaporation onto heated substrates (approx. 400 C) or by DC-sputtering onto

a heated substrate.

There are two important physical methods for the preparation of absorber films, simul-

taneos thermal evaporation of the elements (coevaporation) and sequential evaporation

followed by a sulfurization step. Whenever a metal ratio [Cu]/[In]>1 is offered for the

reaction with sulfur species, binary copper sulfides segregate at the surface of the ab-

sorber films. Highly selective KCN etching solution is used to remove the segregation.

For the deposition of the CBD-CdS buffer layer, CuInS2 films are immersed into an

aqueous solution from NH3, Cd(CH2COO)2 and NH2-CS-NH2 at 60 C for 7 min fol-

lowed by rinsing with deionized water [50]. Finally, the ZnO window layer is grown by

22 Chapter 2. Experimental procedures

RF sputtering. First, a 100 nm thick undoped i-ZnO is deposited followed by a 400 nm

Al doped n+-ZnO layer. The Ni/Al front grids are deposited by e-gun evaporation.

2.1.1 Absorber Preparation

There are a variety of chemical [51] and physical methods [37, 38] for the preparation

of thin CuInS2 films. Device grade materials were up to now exclusively obtained from

the two physical methods: Simultaneous evaporation of the elements and sequential

evaporation of the metals followed by a sulfurization step. A schematic representation

of a vacuum system for simultaneous evaporation is given in Figure 2.2 on the left.

The elements are evaporated from Knudsen cells at 1280 C for Cu, 920 C for In and

S

Cu InSub-strate

Cu InS

Substrate heating

Substrate

Knudsen-cells

a)

b)

Substrate

holder

Substrates

X-tal

balance

Cu-source In-source

Figure 2.2: (Left:) Schematic representation of a vacuum system for the preparation

of thin CuInS2 films by simultaneous evaportation of the elements. (a) Side

view (b) top view. Absorber layers with varying cation ratio can be prepared

in such a system by taking advantage of the displaced evaporation cells.

(Right:) Sketch of a vacuum system used for the sequential evaporation of

the metals. The substrates were rotated for homogeneous deposition.

220 C for S. The substrate temperature is 550 C - 600 C. The spatial configuration

of the evaporation cells opens up the possibility to prepare absorber layers with lateral

varying stoichiometry. Such layers were used in this work to study the effects of varying

stoichiometry on CuInS2 phonon modes.

The sequential preparation process requires an additional sulfurization step in another

vacuum system. For precursor preparation a layer of Cu is deposited on the Glass/Mo

2.1. Solar Cells Based on CuInS2 23

substrate. Subsequently a In layer was added. A sketch of the used vacuum system

is given in Figure 2.2 on the right. The metals are evaporated from electrical heated

tungsten boats. Mass flow is controlled by X-tal balances in a servo loop. Shutter were

used for defined start and stop of the deposition (not drawn). The substrates were

rotated for homogeneous deposition.

The chemical reaction of thin Cu-In films with H2S occurs on a time scale of seconds

[52]. In order to gain more control over reaction kinetics, it is necessary to apply

fast sulfurization processes. From a technological point of view, fast processes are

desirable for high throughput. Laser annealing of elemental precursors was applied

for the synthesis of thin CuInSe2 films [53], but later works concentrated on rapid

thermal processes [7, 6]. Those systems were equipped with halogen lamp arrays. In

the system used for this work, samples can be heated up to 550 C within 0.5 min

by the heat radiation due to their low thermal mass whereas the reactor walls remain

cool. A schematic view is given in Figure 2.3. The temperature at the sample holder

5 % H S in Ar2

Bus

Pyrometer

MFC1 MFC2

N2

Lamp arraySamples

Thermo couple

Sample transfer

Figure 2.3: Schematic view of the used rapid thermal procesing (RTP) system with

mass flow controllers (MFC).

is measured by a pyrometer and by a thermo couple below 200 C. Temperature is

adjusted by a closed servo loop. Contamination of the reactor is avoided by several N2

purge and evacuation cycles prior to sulfurization. Gas flows were kept constant using

mass flow and pressure controllers. CuInS2 films for this work were grown maintaining

a constant gas flow of typically 500 sccm of 5 vol% H2S in Ar during sulfurization

reaction. The total pressure within the reaction chamber was kept at 500 mbar. When


the desired pressure and gas flows were reached, the precursors were heated up to the

preset sulfurization temperature. Heat ramp ∆ T/∆ t can be set as high as 18 C /

sec. Sulfurization time tsulf = 0 refers to the moment when the preset temperature

Tsulf is attained.

2.2 Raman Spectroscopy

In this section fundamentals of inelastic light scattering will be provided. The intention

is to give in brief important concepts of the theory necessary for the interpretation of

the experimental data rather than a comprehensive presentation of Raman scattering.

For the latter a series of textbooks and articles are available [54], [55]. At the end of

the chapter a description of the Raman setup used for this work is provided.

2.2.1 Macroscopic Theory of Inelastic Light Scattering by

Phonons

The interaction of light from the visible spectrum with the solid is intermediate by

the polarizability of the valence electrons. When a electromagnetic field E is present

in the medium, the polarization P will be induced:

P = ε0 χ∼ E. (2.1)

The periodic variation in P is responsible for the emission of a wave which is the

inelastic scattered wave. Within the framework of classical electrodynamics the

scattered light can be described as the oscillation of an ensemble of dipoles. The

Raman scattering intensity can thus be expressed by the dipole radiation intensity

using the transition susceptibility χ∼ :

Is = Iiω4

sV

(4πεε0)2c4|es χ

∼ ei|2 (2.2)

where Ii,s and ei,s denote intensity and polarization unit vector of incident and scattered

light, and V is the scattering volume. The modulation of the susceptibility in general

is caused by collective excitations, i.e. fluctuations in electron density or deflection

of atom cores from their idle states. Phonons can be described as periodic lattice

deformations:

Q = Q0 cos[Ω(q)t]. (2.3)

Expanding χ∼ in a Taylor series and writing the first two terms results in :

χ∼ = χ

∼0 + (∂ χ

∼ /∂Q)Q + ... (2.4)

2.2. Raman Spectroscopy 25

Describing the incident light wave as

E = E0 cos ωit. (2.5)

the polarization (2.1) can be rewritten with (2.4) and (2.5):

P = ε0 χ∼

0E0 cos ωit + ε0

∂ χ∼

∂QQ0 E0 cos [Ω(q)t] cos ωit

= ε0 χ∼

0E0 cos ωit +1

2ε0

∂ χ∼

∂QQ0 E0 cos[ωi + Ω(q)] t + cos[ωi − Ω(q)] t.(2.6)

The first term in (2.6) corresponds to the elastic scattered part of the stray light.

The side bands with frequencies ωi ± Ω expressed in the second term correspond to

the inelastic scattered part. The processes are referred to as Rayleigh- and Raman-

scattering, respectively. The side band with lower frequency is known as Stokes-line

and the one with higher frequency as Anti-Stokes-line.

2.2.2 Raman Tensor and Selection Rules

The first term in (2.4) corresponds to elastic Rayleigh scattering, the second one de-

scribes one-phonon scattering processes. Higher terms in the expansion originate from

one or more phonon amplitudes. The partial derivatives in (2.4) constitute the Raman

polarisability, often termed as Raman tensor R∼ . For a first order one-phonon Raman

process, R∼ is given by the complex second rank tensor

R∼ =

∂ χ∼

∂QQ(ω0) (2.7)

where Q(ω0) is the unit vector of the displacement Q of a given atom. The Raman

scattering intensity

Is ∼| ei · R∼ · es |2 . (2.8)

depends on the polarization of the incident and scattered radiation. By measuring the

dependence of the scattering intensity on the incident and scattered polarization one

can deduce the symmetry of the corresponding Raman active phonon. Symmetries of

the medium and of the vibrations involved in the scattering impose requirements on

the Raman tensor. The result of these symmetry requirements is that the scattered

radiation vanishes for certain choices of the polarisation ei and es and scattering ge-

ometries. This are the so-called Raman selection rules.

The scattering geometry is specified by four vectors ki,ks, ei and es. These four vectors

define the scattering configuration usually represented as ki(ei, es)ks which is known

as the Porto notation.


A compilation of Raman tensors for the 32 crystallographic point groups are presented

in [58]. The Raman tensors for the chalcopyrite lattice are given in Table 2.1.

Table 2.1: Raman tensors and their symmetries for the point group D2d [58].

A1 B1 B2 E, x E, ya 0 0

0 a 0

0 0 b

d 0 0

0 −d 0

0 0 0

0 e 0

e 0 0

0 0 0

0 0 f

0 0 0

g 0 0

0 0 0

0 0 f

0 g 0

2.2.3 Microscopic Theory of Raman Scattering

Light scattering can also be described within quantum mechanical theory. The Raman

process can be virtually decomposed into three electronic transitions [54]:

• the electronic transition from the ground state |0〉 to an excited state |a〉: creation

of an electron-hole pair due to the absorption of a photon with the energy hωi.

• the electron-lattice interaction, i.e. the electronic transition from |a〉 to |a′〉 under

creation or annihilation of a phonon with hωs.

• the transitition from |a′〉 to the ground state |0〉: recombination of the electron-

hole pair under emission of a photon hωs.

For the combination of these processes the third-order pertubation theory yields as the

dominant term for the Raman scattering probability for a given phonon mode

Pph(ωi) ≈ (2π

h)

∣∣∣∣∣〈0|p(ωs)|a′〉〈a′|He−ph|a〉〈a|p(ωi)|0〉(Ea′ − hωs)(Ea − hωi)

+ c

∣∣∣∣∣2

(2.9)

where p(ωi) and p(ωs) are vector components of the dipole operators of the scattered

and incident light, He−ph is the electron-phonon interaction Hamiltonian and c a con-

stant. Ea is the energy of the intermediate state i.e. of an exciton.

Resonant Raman Scattering

If the energy of the incident photons come close to the energy of excited electronic

states in the medium the generation or annihilation of electron-hole pairs increases

dramatically. In consequence an enhanced Raman scattering intensity is observed.

2.2. Raman Spectroscopy 27

This expressed by the denominator in (2.9). The difference between hωi and hωs is

equal to the phonon energy and is usually small compared with electronic energies.

Whenever (Ea − hωi) is small (Ea′ − hωs) will also be small. Thus in (2.9) there are

two resonant denominators. The case Ea = hωi is referred to as an incoming resonance

while Ea′ = hωs is an outgoing resonance.

The intermediate state |a〉 has a finite lifetime τa due to radiative and nonradiative

decay processes. To take account of this fact Ea was expressed by a complex energy

Ea− iΓa, where Γa is the damping constant related to τa by Γa = h/τa. If the resonant

state Ea is a discrete state i.e. an exciton and is well separated from other intermediate

states, the Raman scattering probability can be rewritten as:

Pph(ωi) ≈ (2π

h)

∣∣∣∣∣ 〈0|p(ωs)|a′〉〈a′|He−ph|a〉〈a|p(ωi)|0〉(Ea′ − hωs − iΓa′)(Ea − hωi − iΓa)

∣∣∣∣∣2

. (2.10)

2.2.4 Experimental Setup for Raman Scattering

In Figure 2.4 an outline of the Raman scattering setup used in this work is given. An

Ar+-ion and a Kr+-ion laser were available. For the filtering of the non lasing plasma

emission lines a laser line filter was used for the green 514.5 nm line of the Ar+-ion

laser. The filtering for other wavelengths was achieved by a prism. It was adjusted

such that only the laser line could pass a diaphragm whereas other wavelengths were

refracted out of the optical path. The polarisation of the light was varied with a Fresnel

rhombus. The laser beam was focused onto the sample by the lens of a microscope.

Using a short focal distance lens for so called micro-Raman measurements offers certain

advantages: Scattered light is collected from a large solid angle. This results in an

enhanced sensitivity of the setup. As the light is focused on a spot only 1-2 µm in

diameter spatially resolved measurements can be performed. On the other hand, the

high power density can result in damages of the chemical or structural properties of

the sample.

A single lens is used for focusing the laser onto the sample and collecting the scat-

tered light. This geometry is called backscattering configuration. Other configurations

are possible as well [54]. The light is focused by a lens system onto the entrance slit of

the monochromator. A polarisation analyser was used to define the entrance plane.

Usually the scattered light is 4-6 orders of magnitude weaker than the elastically scat-

tered light. At the same time the difference in frequency between the Raman signal

and the laser light is only about 1 % of the laser frequency [25]. In order to detect this

small sidebands a good spectral resolving power and an excellent stray light rejection

ratio is required.

The Dilor xy-800 Raman spectrometer used for this work was equiped with a double

monochromator for high efficient stray light reduction. In the first monochromator

(M1) the incident light was dispersed. The exit slit was set such that the spectral


Figure 2.4: Raman scattering setup with light source und triple monochromator con-

sisting of a subtractive double monochromator (M1, M2) for effective stray

light reduction and a third monochromator (M3) for the spectral dispersion

[54].

components containing the laser frequency was cut of by the edge of the slit. This

configuration is called the subtractive mode. The spectral separation was reversed by

the second stage (M2) and focused on the entrance slit of the third monochromator

(M3). All of them were equipped with a holographic gratings of 1800 lines/mm. The

spectrum was focused on a CCD camera cooled by liquid nitrogen to reduce the ther-

mal noise. Resolution of the setup depends on the laser frequency. Using the green line

(514.5 nm) of the Ar+-ion laser, an interval of 550 cm−1 was reproduced on 733 diodes

of the camera. This corresponds to 0.75 cm−1 per detector element. The FWHM of

a spectral plasma line was determined to 1.6 cm−1 when the slit was set to 100 µm.

This is the resolution of the setup for the given parameters.

Chapter 3

Raman Spectroscopy of Thin

CuInS2-Films

It will be shown in this work that Raman spectroscopy is a powerful tool for the

analysis of thin CuInS2 films. Choosing appropriate measuring conditions, it is a

destruction-free method, thus the analyzed layers are subsequently available for further

characterization and/or the preparation of solar cells. In order to avoid any damage to

the chemical or structural properties, the power density of the laser must not exceed a

certain threshold. This threshold was determined for CuInS2 layers in Section 3.1.

Due to the polycrystalline character of the absorber films vibrational properties may

vary laterally. This was studied with the micro-Raman technique and a Raman map

will be presented in Section 3.2. It will be shown that the morphology of the absorber

is reflected in the Raman map.

Resonant Raman scattering at fundamental band gap energies was performed using a

tunable laser. The scattering cross section contains information about the electron-

phonon interaction involved in the process. The resonance curves of CuInS2-films were

determined and will be presented in Section 3.3. The occurrence and the width of the

resonance curves were highly dependent on the crystal quality of the layers. Narrow

resonance curves were observed from high quality layers. The involvement of a free

exciton responsible for the increase in scattering intensity will be discussed. It will be

concluded that resonant Raman spectroscopy of is a sensible tool for the determination

of CuInS2 crystal quality.

Dependency of the Raman cross section on photon energies above the fundamental

band gap will be discussed in the last section of this chapter.

29

30 Chapter 3. Raman Spectroscopy of Thin CuInS2-Films

3.1 Laser Induced Defects

Most Raman measurements performed in this work were carried out on a micro-Raman

setup. This technique offers a number of advantages over a macro setup (see Section

2.2.4). The incident laser light is focused within a spot of 2 µm in diameter on the

sample by a short focal distance lens. In this manner a high power density is achieved.

Moreover, the scattered light is collected from a large solid angle. In summary an

enhanced sensitivity of the setup is realized. On the other hand the high power density

can damage the chemical or structural properties of the material. This is especially

crucial for CuInS2 as the absorption coefficient of the material is very high (see Section

1.2.2) and local heating by the laser light is expected. Non-destructive heating of the

layer causes a shift in phonon frequencies and broadening of line widths. Both changes

are functions of the temperature and the changes are reversible. Lattice or chemical

damages may be indicated by the same changes in the spectral features, but they are

not reversible. This difference can be exploited to determine whether the laser induced

a non-destructive thermal effect or damaged the layer.

In order to find the critical power density, from which on damages will be observed, a

series of measurements with increasing laser power was performed. First, the CuInS2

film was irradiated by the laser for 10 minutes with the power density in question.

Down cooling of the irradiated spot was made possible during a break of 15 minutes.

Then, all spectra were recorded at a moderate power density Plaser = 4 kW/cm2.

Each measurement was performed on a different, previously non-irradiated spot on the

sample. The A1-phonon mode of two spectra of the series is given on the left panel in

Figure 3.1.

3.2. Lateral Inhomogeneities of Polycrystalline CuInS2 Films 31

270 280 290 300 310

1

0

After laser irradiation with

Plaser

= 4 kW/cm2

Plaser

= 40 kW/cm2

Nor

mal

ized

Int

ensi

ty

Raman Shift (cm-1)

0 10 20 30 40289,0

289,5

290,0

290,5

291,0

291,5

292,0

292,5

293,0

Intensity

Ram

an I

nten

sity

(ct

s /

mW

⋅min

)

FWHM

FW

HM

(cm

-1)

Ram

an S

hift

(cm

-1)

Power Density (kW/cm2)

6,0

6,5

7,0

7,5

8,0

8,5Threshold

RamanShift

16

20

24

28

32

36

40

44

Figure 3.1: Study of laser induced structural defects. A polycrystalline CuInS2 film was

irradiated with varying laser power densities (4 -40 kW/cm2, λ=514.5 nm).

After a cooling down break Raman spectra were recorded with moderate laser

power density. (Left:) A1 phonon mode recorded after laser irradiation with

Plaser=4 and 40 kW/cm2, respectively. (Right:) Development of the shift in

phonon frequency, FWHM and scattering intensity over the power density

of the laser treatment.

After laser irradiation with Plaser=40 kW/cm2, an increase in peak position at

2.3 cm−1 and a difference in the full width at half maximum (FWHM) of 2.7 cm−1

was observed. This was clearly indicating a damage to the layer. This values were

evaluated for every A1-mode from all the spectra. They are shown on the right panel

in Figure 3.1 in dependency on the laser irradiation. No changes were observed up to

16 kW/cm2. Then scattering intensity started to decrease, line widths was broadening

and the phonon mode started to shift towards lower frequencies. In view of this results

all the measurements presented in this work were recorded with laser intensities around

8 kW/cm2 (λ = 514.5 nm) or less.

3.2 Lateral Inhomogeneities of Polycrystalline

CuInS2 Films

The information derived from micro-Raman measurements are local information as the

laser spot is about 2 µm in diameter. This corresponds to the size of the crystallites

which are about 2-3 µm. Lateral inhomogeneities and local defects may be reflected in

micro-Raman spectra taken from different spots on the sample. Therefore comparison

of a single spectrum with another one from the same or another sample is only valid


within certain limits. Lateral variations in Raman data will be the discussed in this

Section.

The grains of the absorber layer may contain dislocations which divide the grain into

small undistorted domains. Such domains increase the Raman mode frequencies and

give rise to broadening of the Raman lines accompanied by a decrease in scattering

intensity [59]. A decrease in the scattering intensity can also be caused by the surface

roughness of the sample. Parts of the incident light can be back scattered from tilt

planes within the laser spot area. This parts may be not collected resulting in sub-

stantial changes in the scattering intensity. However, only scattering intensities are

affected whereas the spectral position of the lines and the FWHM are conserved.

A single phase CuInS2 film was mounted on a manually driven xy-stage which is part

of the microscope used for focusing the laser and collecting the scattered light (see

Section 2.2.4). As the microscope was equipped with a crosshair, it was possible to

control the manipulation of the stage position in the micron range. A central spot was

chosen on the surface and spectra were recorded from matrix points within a 10x10 µm

square. As the A1-mode of CuInS2 is the most intense mode it was analyzed for each

spot. Scattering intensity and full width at half maximum (FWHM) were encoded in

a color scheme ranging from black to white. The matrix is depicted in Figure 3.2.

Figure 3.2: (Right:) Raman frequency mapping of the A1-phonon mode, (middle:)

scattering intensity and (left:) full width at half maximum. The sample

was a single phase polycrystalline CuInS2 absorber layer.

Different scattering intensities and FWHM-values were obtained from adjacent

spots. By comparing the two matrices a correlation of the FWHM-values and the

scattering intensities can be recognized even the correlation is not perfect. This might

3.3. Resonant Raman Scattering 33

be due to additional variations in the scattering intensity caused by the surface rough-

ness as explained above.

It can be concluded that local variations in measures taken from the Raman spectra

are governed by the polycrystalline character of the layer. The maximum variations

from a single randomly chosen spot in comparison with another spot on the same sam-

ple was ± 0.5 cm−1 and ± 2.5 cm−1 for the Raman shift and FWHM, repectively.

When calculating the mean value and the standard deviation for data from 5 spots the

standard deviation was 0.2 cm−1 for the Raman Shift and 0.9 cm−1 for the FWHM.

Measurements for this work were therefore taken from at least three up to five different

spots on the sample and mean values were calculated.

3.3 Resonant Raman Scattering

Raman scattering intensity is a function of the excitation energy as was shown in Sec-

tion 2.2.3. An increase in intensity can be observed at energies close to the band gap

where the joint density of states is high. The resonance behaviour of polycrystalline

CuInS2 samples was studied and will be the subject of this section.

Absorber layers used for the resonant Raman experiments were prepared by the se-

quential process. The metal atom ratios of the precursors were [Cu]/[In]=1.2 and

[Cu]/[In]=0.8. The stacks were etched after sulfurization (Tsulf=525 C, tsulf=10 min).

It will be shown later that this procedure results in stoichiometric CuInS2 layers in case

of Cu-rich precursors. Defect rich absorbers are expected in the case of Cu-poor pre-

cursors. For the resonant Raman measurements another setup than the one presented

in Chapter 2.4 was used. The second setup was equipped with a tunable Ti:sapphir

laser pumped by the 514.5 nm line of an Ar+ ion laser. A convex lens was used to

focus the laser light onto an area of 100 µm in diameter onto the sample. Considering

the size of the light spot, the setup must be termed as a macro-setup. Overall laser

power on the sample was 20 mW, resulting in 255 W/cm2 power density in the laser

focus.

Raman spectra and the intensity of the A1-phonon mode versus excitation energy of

a Cu-rich prepared sample are shown in Figure 3.3. Maximum enhancement was ob-

served at 1.506 eV. The ratio of the lowest A1-phonon mode intensity at Eex=1.580

eV to the maximum mode intensity at Eex=1.506 eV was 3.6. No other features were

observed in the spectra besides the A1 mode. In contrast, for the Cu-poor sample no

resonance behaviour was observed. Only a weak structure at the position of the A1-

mode was present. The count rate of the signal was just little above the background

noise and disappeared completely below 1.5 eV.

The sharp resonance curve of CuInS2 can be explained by the involvement of bound

exciton states at the band edge (refer to Section 1.2.3). The band gap at 300 K is 1.535

eV [14]. The free exciton was found 20 meV below the band gap at 1.515 eV and a


240 260 280 300 320 340

200

291 cm-1

A1

Eex = 1.501 eVEex = 1.494 eVEex = 1.490 eVEex = 1.473 eV

Inte

nsity

(ct

s / m

W*m

in)

Raman Shift (cm-1)

780 790 800 810 820 830 840 850

2

4

6

8

10

Excitation Energy (eV)

Inte

ns

ity

(C

ts /

mW

*min

)Excitation W avelength (nm)

1.58 1.56 1.54 1.52 1.50 1.48 1.46

2

4

6

8

10

Figure 3.3: (Right:) Raman spectra of a Cu-rich prepared CuInS2 sample at different

excitation energies recorded at room temperature and (Left:) Intensity of

the A1-mode of over excitation energy. The dashed line is a fit to the data

points using equation (3.1).

bound excitons at 1.500 eV (Section 1.2.3). A single resonance peak was found thus

the dependence of the Raman cross section σi on excitation energy due to an incoming

resonance can be described by using equation (2.9)

σi∼=

α

(Ea − hωi)2 + Γ2a

(3.1)

where Ea and Γa are energy and damping constant of an intermediate state, respec-

tively, and α is a constant. The term hωi is the energy of the incident photons. Data

points in Figure 3.3 were fitted using equation (3.1). The damping constant corre-

sponds to the half width at full maximum (HWFM) and was determined from the fit

to be Γa = 22 meV.

Wakita et al. observed a strong enhancement of the Raman modes from a CuInS2

single crystal [60]. The FWHM of the resonance curve there was 3 meV at 9 K. The

resonance was attributed to a bound exciton at 1.525 eV. The energy shift of the

bound exciton was assumed to be in the order of the shift in band gap energy when

temperature is increasing from 9 K to room temperature. The band gap energy shift is

∆EG= -20 meV [14]. Additionally, the binding energy of the excitons can vary slightly,

depending on the typ of defect they are bound to. Taking this facts into account, the

results are quit in accordance. The broader resonance curve found for the polycrys-

talline samples can be attributed to the higher level of defects of this material. In

consequence the resonance curve is broadening. This interpretation is supported by

3.3. Resonant Raman Scattering 35

the fact that no resonance behaviour was found for the Cu-poor sample were the defect

density is much higher than in Cu-rich samples [31]. Alvarez et al. [61] observed a

similar behaviour from polycrystalline CuInS2 samples prepared by coevaporation. He

found an enhancement of the A1-mode by a factor 4 and the maximum of the resonance

curve was at 1.47±0.02 eV which is in agreement with the results presented here for

the samples prepared by the sequential process. Alvarez found no resonance behaviour

from Cu-poor samples, too, and attributed this finding to the poor crystal quality of

the those samples.

Summarizing the own results and the reports from the literature, it can be established

that the resonance effect is changing dramatically with the crystal quality of the layer.

They range of the effect spans the absence of any resonance in the case of Cu-poor

polycrystalline CuInS2 films up to extremely narrow resonance curves observed for sin-

gle crystals. Therefore the resonance behaviour is a sensitive indicator for the crystal

quality of of CuInS2 films.

3.3.1 Raman Scattering at Excitation Energies above the Fun-

damental Gap

Raman scattering intensities above the fundamental band gap depend mainly on the

electronic density of states for interband transitions. The measured scattering intensi-

ties depend additionally on the sensitivity of the Raman setup in use for a given wave-

length and polarization. Scattering intensities were measured for the three strongest

available frequencies of the Ar+-ion laser. The blue line (2.708 eV), the green line

(2.410 eV) and the red line (1.916 eV) of the Ar+-ion laser were used. The polarization

of the laser light was parallel to the entrance slit. A polycrystalline single phase CuInS2

sample was used for the measurements. The spectra are given in Figure 3.4.

The A1-mode is the dominant feature in the spectra. Three more peaks were observed

and assigned to known phonon modes. They were a magnitude lower in intensity than

the A1-mode and barely above the noise.

The measured scattering intensities for the A1-mode was slightly lower for the green

laser light and over a magnitude lower for the red light in comparison to the blue light.

This was due to spectral and spatial sensitivity of the Raman setup (refer to Section

2.2.4). Measured scattering intensities are compiled in Table 3.1 together with the

sensitivity of the setup. Intensities were in the same order of magnitude after nor-

malization to 100 % sensitivity. The remaining differences can be explained by the

influence of the varying density of electronic states in the spectral regions.

The green laser line of the available Ar+-ion line was by far the most intense. In view

of the satisfying sensitivity of the Raman setup in this spectral region, the green laser

line (514.5 nm) was chosen for comprehensive analysis of CuInS2 layers prepared in

this work.


150 200 250 300 350 400

35

03

40

32

6

2932

66

24

4

500

Eex

= 2.708 eVE

ex = 2.410 eV

Eex

= 1.916 eV

Inte

nsity

(ct

s /

mW

⋅ min

)

Raman Shift (cm-1)

Observed Lit.[10] Symmetry

(cm−1) (cm−1)

244 244 E3TO

266 260 E3LO

293 294∗ A1

326 323 B12 TO

340 339 E1LO

350 352 B12 LO

Figure 3.4: Raman spectra of a standard CuInS2 sample at different excitation ener-

gies above the band gap. In the table: Observed phonon frequencies and

literature data for comparison (in cm−1). Asterisks indicate modes which

are measured at T = 78 K.

Table 3.1: Measured Raman intensities for the CuInS2 chalcopyrite A1-phonon mode

for different excitation energies. Intensities were corrected for instrumental

throughput in the last column.

Excitation Raman Corrected

energy intensity intensity

(eV) (cts/mW·min) (cts/mW·min)

2.708 1400 1600

2.410 1300 7000

1.916 80 4000

Chapter 4

Non-Chalcopyrite Phonon Modes in

CuInS2

Thin CuInS2 layers were prepared by sequential physical vapor deposition of the metals

and subsequent sulfurization in H2S. The dependence of the structural properties on

sulfurization parameters and composition was analyzed by Raman spectroscopy. Those

studies revealed modes at 305 cm−1 and 60 cm−1 in the spectrum of the sulfurized layers

that could not be assigned to known phonon modes from compounds appearing in the

(Cu-In-S)-system. In this thesis it will be shown for the first time that those modes

can be attributed to the CuInS2 CuAu ordered phase.

Raman spectra showing the non-chalcopyrite modes will be presented in the first section

and the spectral features will be discussed. The CuInS2 CuAu ordered structure as an

origin for those modes will be formulated as a working hypothesis. Raman selection

rules for the CuAu phonon mode at 305 cm−1 were determined by own measurements.

It will be shown in Section 4.1.2 that this mode was totally symmetric. The number of

phonon modes, their symmetry and optical activity for given crystal structures can be

determined by group theory. Only the number of elements, the sites of the elements in

the unit cell and the symmetry of the crystal structure are inputs to these calculation.

Thus group theory is a powerful tool to predict phonon properties for a given structure.

This tool was applied for the CuAu structure and the results are given in Section 4.1.3.

It was found, that a single totally symmetric phonon mode should be observed from the

CuAu structure. This result supported the assumption of CuAu ordered domains in

the measured films and encouraged the performance of phonon frequency calculations

from first principles.

Phonon frequencies for the CuInS2 CuAu structure were calculated for the first time

in this work. The calculated frequencies matched the observed phonon frequencies.

Thus the calculations confirmed the assumption of CuAu ordered domains present

in the analyzed absorber layers. The appearance of the CuAu ordered domains was

studied in dependence on the preparation conditions. It will be shown in Section 4.2

37

38 Chapter 4. Non-Chalcopyrite Phonon Modes in CuInS2

that CuAu ordered domains can be related to CuInS2 layers prepared under Cu-poor

conditions and to layers sulfurized in H2S in a low temperature regime. The preparation

conditions that lead to the formation of layers with CuAu ordered domains coincide

with those that lead to solar cells with poor conversion efficiency. This relation will

be discussed and a tentative explanation will be provided. The presence or absence of

phonon modes from CuAu ordered domains can be taken as a quality indicator for the

prepared absorber layers. This opens up the possibility to use Raman spectroscopy as

a tool for ex- and in-situ growth monitoring of absorber layers.

4.1 Observation of CuAu Order and Theoretical

Considerations

4.1.1 Observation of Non-Chalcopyrite Phonon Modes

Non-chalcopyrite phonon modes were observed from a variety of thin CuInS2 films

prepared by different techniques and parameters. They were found in films prepared by

the sequential process, by co-evaporation and by MOCVD growth when the cation ratio

of the film was [Cu]/[In]<1. Furthermore those phonon modes were found when Cu-In

stacks were sulfurized in H2S at low temperatures (375 - 475 C). Details from own

measurements will be provided later. A typical spectrum recorded from Cu-deficient

absorber layer is shown in Figure 4.1.

100 200 300 400 5000

100

200

300

400

500

600

700

800

259

242

294

305

340

60

Raman Shift (cm-1)

Inte

nsi

ty (

cts/

mW

⋅min

)

Comp. Obs. Lit. Symm. Ref.

CuInS2: 242 244 E3TO [10]

Chalcopyrite 259 260 E3LO [10]

294 294∗ A1 [10]

340 339 E1LO [10]

CuAu ?? 60 n.o. ?

305 309 ? [62]

Figure 4.1: (Left:) Raman spectrum of a thin CuInS2 layer. The composition of the

film was [Cu]/[In]=0.8. The modes at 60 cm−1 and 305 cm−1 do not belong

to the CuInS2 phonon spectrum. It will be shown in this work that those

modes can be attributed to the CuInS2 CuAu-ordered structure. (Right:)

Assignment of the observed phonon modes and literature data. (*) at 80 K.

4.1. Observation of CuAu Order and Theoretical Considerations 39

Due to the elastic scattered light of the laser, an continued increase in Raman in-

tensity was measured at frequencies lower than 50 cm−1. The laser was focused by a

lens of short focal length in the micro-Raman setup. Thus, there was a high power

density in a small air volume close to the sample surface. On a first glance, the spec-

trum in Figure 4.1 seems to be noisy below 150 cm−1. The noise can be identified as

contributions from close spaced, narrow peaks, low in intensity. They originate from

rotational vibrations of nitrogen molecules [58]. However, in most cases, the mode at

60 cm−1 did appear together with the mode at 305 cm−1 for CuInS2 layers and was

therefore not attributed to nitrogen vibrations.

The observed phonon modes were assigned to known CuInS2 chalcopyrite phonon

modes by comparison with literature data, except the two modes at 305 cm−1 and

60 cm−1. In literature, a phonon mode at 309 cm−1 from a Cu-poor polycrystalline

CuInS2 film was first reported by Morell [62]. Hunger et al. [63] reported a mode at

309 cm−1 from epitaxial CuInS2 layers grown on Si(111). The films were Cu-poor and

the mode at 309 cm−1 was the most intense. The chalcopyrite A1-mode was present at

the same time but a magnitude lower in intensity. Both authors suggested the presence

of spinel-structured β-In2S3 in the layers. The hint for this compound was a mode at

306 cm−1 observed by Kampas et al. [64] in the Raman spectra of β-In2S3 crystals.

He reported fifteen Raman active modes from β-In2S3 crystals. Modes at 70 cm−1 and

244 cm−1 were more intense that the one at 306 cm−1. The absence of strong phonon

contributions from Cu-poor CuInS2 layers made the proposed presence of β-In2S3 less

likely. Kondo et. al [12] related the phonon mode at 307 cm−1 to polycrystalline sam-

ples exhibiting a poor crystalline quality. He suggested a localized mode with a smaller

mean atomic weight of the cations. Alvarez et. al [13] pointed out that changes in the

arrangement of Cu and In atoms in the cation sublattice could be responsible for the

observation of the phonon mode at 307 cm−1. The involvement of sphalerite, a phase

of CuInS2 and the CuAu structure were discussed. The presence of sphalerite was later

ruled out due to the negative results of XRD measurements .

In this work the phonon mode at 305 cm−1 will be attributed to the co-existence of

CuAu ordered CuInS2 domains along chalcopyrite domains in those films.

4.1.2 Raman Selection Rules

In order to learn more about the nature of the phonon mode at 305 cm−1, Raman se-

lection rules were determined. For this purpose epitaxial CuInS2 layers on Si(111) were

analyzed. The samples were Cu-poor, the cation ratio was nominally [Cu]/[In]= 0.85.

Measurements were performed in the experimental geometric configurations 〈z|xx|z〉,〈z|xy|z〉, 〈z|yx|z〉 and 〈z|yy|z〉 where x, y and z correspond to the crystallographic

directions [112], [110] and [111] of the silicon substrate. Raman spectra are given in

Figure 4.2. The chalcopyrite phonon mode at 294 cm−1 behaves as expected from a


100 150 200 250 300 350 400 450

294

<z|xy|z>10

0

307

<z|xx|z>Inte

nsity

(ct

s /

mW

⋅ min

)

Raman Shift (cm-1)

Figure 4.2: Raman spectra from a epitaxial CuInS2 layer measured in two different

polarization configurations (300 K). Wavelength of the incident laser light

was λ=514.5 nm. This work.

mode with A1 symmetry: It is observed in parallel configurations 〈z|xx|z〉, 〈z|yy|z〉 and

a dramatic decrease in intensity was found in the Raman forbidden 〈z|xy|z〉, 〈z|xy|z〉configurations. The mode at 305 cm−1 behaves the same way: It was observed together

with the A1-chalcopyrite mode in parallel configurations and decreased in the same ra-

tio as the A1-chalcopyrite mode. From this finding it can be concluded that the mode

at 305 cm−1 has the same symmetry: A1.

4.1.3 Group Theoretical Analysis

The analyzed absorber layers exhibited two totally symmetric phonon modes. The

CuInS2 chalcopyrite structure contributes one of them thus the second one could be

contributed from the CuAu-ordered structure. The number of Raman active phonon

modes and their symmetry can be calculated with group theoretical methods. Only

the symmetry of the structure has to be provided as an input. Group theoretical anal-

ysis for the CuAu-like structure was not found in the literature, therefore the analysis

was performed in this work and will be presented here. The fundamentals of group

theory and crystallography will not be discussed in this work. Fundamental concepts

of crystal symmetries are given in [17] and [65]. A good introduction to abstract group

theory can be found in [66] and examples for the application of group theory to solid

state problems are given in [67] and [68]. To calculate the number of normal modes

and their symmetry the method after Porto [69] was applied. The calculation was per-

formed for the CuInS2 chalcopyrite structure to demonstrate the procedure and then

for the CuAu structure. The following four steps were performed:


First step: The Wyckoff positions for the three elements in the unit cell must be deter-

mined for the given space groups. D122d and D5

2d are the space groups for chalcopyrites

and the CuAu-like structure, respectively.

Table 4.1: Chalcopyrite belongs to the space group D122d and the CuAu ordered structure

to the space group D52d. All sites for the two space groups are given [69].

Space Group Sites

D52d(P 4m2) ∞[lC1(8)] +∞[(k + j)CS(4)] +∞[(i + h)C2(4)]

+∞(g + f + e)C2v(2)] + (d + c + b + a)D2d(1)

D122d(I 42d) ∞[eC1(16)] +∞[dC2(8)] + (b + a)S4

In the chalcopyrite structure four AI , four BIII and eight CV I ions can be accom-

modated. All possible sites for the two space groups are given in Table Table 4.1. By

inspection it can be recognized that AI and BIII ions occupy the a+b sites and CV I

ions must be at the d site since this is the only one which can generate eight equivalent

positions.

Second step: The irreducible representation for all possible lattice vibrations must

be determined. The contribution of each possible site in the space group is given in

Table 4.2.

The representation for all chalcopyrite modes is given by the direct sum of the

representations of the contributing sites:

Γtot = (A1

⊕2A2

⊕B1

⊕2B2

⊕3E)

⊕2(B1

⊕B2

⊕2E)

= A1

⊕2A2

⊕3B1

⊕4B2

⊕7E . (4.1)

Third step: The number of normal modes must be determined. In each mode of

normal vibration all the atoms in the structure vibrate with the same frequency and

all atoms pass through their equilibrium positions simultaneously. This corresponds

to k = 0. For the determination of normal modes the acoustic phonons must be

subtracted since they are propagating waves. Acoustic phonons can be described using

translation vectors. The representations of the translations Tx, Ty, Tz can be obtained

by inspection of the character Table 4.3.

Tx and Ty correspond to the double degenerate mode E and Tz to the B2 symmetry.

Thus the representation for the acoustic modes can be written as:


Table 4.2: Irreducible representations that result from occupying each of the sites within

the space groups Dx2h [69].

Site Representations

C1 3A1⊕

3A2⊕

3B1⊕

3B2⊕

6E

Cz2 (Cz

2 ) A1⊕

A2⊕

B1⊕

B2⊕

4E

C2(C2) A1⊕

2A2⊕

B1⊕

2B2⊕

3E

CS 2A1⊕

A2⊕

B1⊕

2B2⊕

3E

D2 A2⊕

B2⊕

2E

C2v A1⊕

B2⊕

2E

S4 B1⊕

B2⊕

2E

D2d B2⊕

E

Γac = B2

⊕E . (4.2)

In summary the normal modes of the chalcopyrite lattice are given by:

Γnormal = Γtot − Γac = A1

⊕2A2

⊕3B1

⊕3B2

⊕6E . (4.3)

Fourth step: The modes in equation (4.3) can be divided into Raman active, infrared

active and silent modes. A mode will be Raman active if the normal mode belongs

to the same representations as quadratic terms of Cartesian coordinates (x2, y2, z2),

their sums and differences (i.e. x2-y2, etc.) or products of Cartesian coordinates (i.e.

xy, yz, etc.). A mode will be infrared active if the normal mode belongs to the same

representation as the translations Tx, Ty, Tz. If there are no coordinates belonging to a

representation, the related mode is silent. The representations for translations are B2

and E as can be taken from column three of Table 4.3. The corresponding modes are

infrared active. Quadratic functions of the coordinates and Cartesian products can be

found in the fourth column. All of the modes are therefore Raman active, apart from

the A2-mode which is a silent mode. In result 13 optical active modes can be observed:


Table 4.3: Character table for the point group D2d [69].

D2d E 2S4 C2 2C′2 2σd

A1 1 1 1 1 1 x2 + y2; z2

A2 1 1 1 -1 -1 Rz

B1 1 -1 1 1 -1 x2 − y2

B2 1 -1 1 -1 1 Tz xy

E 2 0 -2 0 0 (Tx, Ty); (Rx, Ry) (xz,yz)

Γopt = A1(R)⊕

3B1(R)⊕

3B2(IR; R)⊕

6E(IR; R) . (4.4)

The same procedure will now be applied for the CuAu structure. The Wyckoff

positions were provided in Table 1.1. The lattice of the unit cell (Fig. 1.3) is occupied

by two CV I ions at the site g and by one AI and BIII ions at the sites a and c,

respectively. The direct sum of the representations from contributing sites are denoted

by:

Γtot = (A1

⊕B2

⊕2E)

⊕2(B2

⊕E)

= A1

⊕3B2

⊕4E . (4.5)

After substraction of the acoustical modes (4.2) the normal vibrations were

obtained:

Γnormal = Γtot − Γac = A1

⊕2B2

⊕3E . (4.6)

By inspection of the characters in Table 4.3 it was found that all of them are Raman

active. Furthermore B2 and E modes are infrared active. Hence in the CuAu structure

6 optical active modes can be observed:

Γopt = A1(R)⊕

2B2(IR; R)⊕

3E(IR; R) . (4.7)

From this results it is expected that CuInS2 layers consisting of domains in chal-

copyrite and CuAu order exhibit two modes with A1 symmetry in Raman spectra.

This result strongly supports the assumption of the coexistence of both structures and

was the motivation to perform phonon frequency calculations for the CuAu structure.


4.1.4 Lattice Dynamics of the CuInS2 - CuAu Structure

Phonon frequencies of the CuInS2 CuAu structure were calculated ab-initio from first

principles. The calculations were performed by Dr. J. Fritsch, Technical University

Regensburg in the framework of this thesis. In order to evaluate the results, the

frequencies of the CuInS2 chalcopyrite structure were calculated for comparison. The

calculations were done to verify the coexistence of both structures in samples which

exhibit the phonon mode at 305 cm−1. A detailed presentation of the theoretical

models implied for those calculations is given in [71]. In the next paragraph a short

outline of the theoretical concepts is given. This is followed by the presentation of the

numerical results and their discussion.

The lattice dynamics of a periodic crystal structure can be described in a harmonic

approximation model [72]. In this model a central potential is assumed and only

quadratic dependencies on the displacements of nuclei are allowed as contributions

to the total energy of the crystal. The second derivative of the potential after the

coordinates of the interacting nuclei correspond to the force constants of the harmonic

oscillator. The displacement of nuclei in the periodic crystal lattice can be described

by a plane wave defined exclusively at the positions of the lattice atoms. The dynamic

equations of the crystal result in a eigen-value problem of the form

Du(q) = Ω2(q)u(q) (4.8)

where D is the dynamical matrix, Ω(q) the frequency of the phonons with wave vector

q and u(q) the amplitude of the plane wave. The dynamical matrix D contains the

force constants which must be calculated in order to solve equation (4.8). This requires

the knowledge of the total energy of the system in a given state. For the calculation

of the total energy it is necessary to solve the quantum mechanical many body prob-

lem including all electrons and nuclei. Computing the general many body problem is

impossible due to the giant amount of parameters involved. But the complexity of the

problem can be reduced with some simplifications and approximations such that it can

be calculated with numerical methods [71]:

1. The mass of the electron is much smaller than the mass of the nuclei. It can be

assumed that the electrons follow the motion of the nuclei instantaneously. Thus

it is adequate to calculate only the electronic ground state for a given configu-

ration of the atoms. This is referred to as the adiabatic or Born-Oppenheimer

approximation.

2. The electronic system can be divided into valence electrons involved in chemical

bonds and core electrons. Pseudo potentials can be constructed such that en-

ergy eigen-values and wave functions match the real eigen-states of the isolated


atom. A well constructed pseudo potential reproduces all important electronic

properties of the solid.

3. In density functional theory the electron density n(r) is used instead of the wave

functions of the electrons for the calculation of their energy. The electron density

is mapped on the ground state wave functions. The ground state energy is ex-

pressed by a functional E[n[r]] which is minimal when the electron density n(r)

corresponds to the ground state density. The Kohn-Sham representation [73] of

the E[n[r]] functional can be efficiently used for computations.

The dynamical matrix D was calculated with the ”frozen-phonon” method. Each

atom was displaced by a small amount from its idle state. Energy differences and

forces were calculated within the density functional formalism. For the calculation of

phonon frequencies the dynamical matrix D was diagonalized in order to solve the

eigen-value problem (4.8). Macroscopic polarizations leading to LO-TO splitting of

phonon frequencies due to Frohlich interaction were not taken into account. Thus the

calculated values were the frequencies of the degenerate TO phonon modes. Frequencies

were systematically to low due to the implemented approximations. They were scaled

by a factor 1.05 to correct this deviation. The results are summarized in Table 4.4.

The calculated frequencies for the chalcopyrite phonon modes and their symmetries

are given on the left together with the experimental values. In a first approach the

calculated phonon frequencies match the experimental values. Thus the theoretical

model was able to describe the chalcopyrite phonon modes. Reasonable results are

expected from the calculation of the CuAu phonon spectrum due to the close relation

of chalcopyrite and CuAu structure. The calculated phonon frequency for the A1 CuAu-

mode matches the observed value well and the frequency match of the weaker E3-mode

is satisfying. This findings once more support the assumption of CuAu coexistence in

CuInS2 chalcopyrite films.


Table 4.4: First columns: Symmetry of the phonon modes. Second columns: Calcu-

lated phonon frequencies for CuInS2 chalcopyrite and CuAu-like structure in

cm−1 [70]. Third columns: Experimental values for CuInS2 [10]. Asterisks

indicate values measured at 80 K. Values for the CuAu-like structure are

from this work. All values were calculated for TO phonons only.

Symm. Calc. Exp. Diff.

A1 285 294∗ -9

B11 318 n.o. -

B21 161 n.o. -

B31 96 n.o. -

B12 316 323 -7

B22 241 234 7

B32 81 79∗ 2

E1 307 321 -14

E2 293 295 -2

E3 250 244 -6

E4 146 140∗ -6

E5 89 88 -1

E6 76 67∗ -9

CuInS2: Chalcopyrite structure

Symm. Calc. Exp. Diff.

A1 305 305 0

B12 287 n.o. -

B22 145 n.o. -

E1 299 n.o. -

E2 236 n.o. -

E3 69 60 9

CuInS2: CuAu-like structure

4.2 Dependence on Stoichiometry

The structural and electronic properties of thin CuInS2 layers depend strongly on the

preparation conditions. The amounts of elements offered for the chemical reaction i.e.

govern the resulting stoichiometry. To study the dependence of structural properties

on the stoichiometry, a sample with a lateral variation in composition was analyzed

with micro-Raman spectroscopy. The CuInS2 films was prepared by multisource phys-

ical vacuum evaporation (coevaporation). Sample holder rotation was turned off to

achieve an inhomogeneous distribution of the elements. The element concentrations

were quantified in dependence on sample position using energy dispersive X-ray (EDX).

The result of the EDX analysis is given in Figure 4.3. The sample was Cu-rich at x=0

mm and was changing to In-rich at x=40 mm. An area close to stoichiometry was

present approximately between x=17 mm and x=21 mm.

Raman spectra were recorded every 2 mm along the layer. The spectra are given in

Figure 4.4. For better clarity a selection of six spectra was chosen. When a CuInS2 ab-

sorber layer is prepared with [Cu]/[In]>1 the excess copper forms a binary CuxS layer

at the top of the absorber. The CuxS layer can be completely removed using a KCN

4.2. Dependence on Stoichiometry 47

0 5 10 15 20 25 30 35 4020

21

22

23

24

25

26

27

28

29 EDX: Cu In S/2

Ele

men

t co

ncen

trat

ion

(at.

%)

x-Position (mm)

Figure 4.3: Lateral variation of element concentrations measured with energy dispersive

X-ray. The CuInS2 film was prepared on purpose with varying cation ratio.

The sulfur concentration was plotted using factor 0.5. The solid lines give

the trend of the data.

etch procedure (see Section 5.3). This is a well known fact, thus it is not surprising to

detect vibrational modes of a CuxS compound in the above spectra. The modes were

assigned to CuS according to the report provided in [74]. The subject of CuxS will be

discussed in detail in Section 5.3.

The phonon mode at 294 cm−1 was assigned to the A1-mode of CuInS2 chalcopyrite

and the modes at 60 cm−1 and 305 cm−1 to CuAu according to the already discussed

results in this chapter. In the upper right panel of Figure 4.4 the intensity development

of the 305 cm−1 mode relative to the A1 mode is depicted. There are no CuAu modes

in the spectra from the region where the layer was Cu-rich. Excess copper formed

CuxS segregations at the surface and the absorber itself was stoichiometric in compo-

sition. Leaving the stoichiometric area towards the Cu-deficient region, the 305 cm−1

mode appeared and was steadily increasing in intensity relative to the mode at 294

cm−1. It can be concluded that the appearance of the mode at 305 cm−1 is related to

Cu-deficient CuInS2 layers.


100 200 300 400 500

[Cu]

/[In]

60

267142

Cu-

poor

Cu-

rich

0.81

0.82

0.88

0.87

1.00

1.41

50

0

47430529463

Inte

nsi

ty (

cts/

mW

⋅ min

)

Raman Shift (cm-1)

0.8 0.9 1.0 1.1 1.2 1.3 1.4

0.0

0.5

1.0

1.5

[Cu]/[In] Cu-richCu-poor

Inte

ns

ity

ra

tio

: I(

30

5)/

I(2

92

)

Comp. Observ. Lit. Ref.

CuxS 63 62 [74]

” 142 142

” 267 267

” 474 475

CuInS2: 294 ∗294 [9]

Ch

CuInS2: 60 60 This

CuAu 305 307 work

Figure 4.4: (Left:) Raman spectra of a thin CuInS2 layer with varying cation ratio

(refer to Fig. 4.3). The wavelength of the incident laser light was λ=514.5

nm and the spectra were recorded at 300 K. (Upper right:) Intensity ratio

of the modes at 294 cm−1 and 305 −1. The solid line is a guide to the

eye. (Lower right:) Assignment of the observed phonon modes. Vibrational

frequencies are given in cm−1. Literature data were measured at 300 K and

80 K (asterisk), respectively. For the origin of the mode at 305 cm−1 see

the text.

Analyzing peak positions and peak widths gave additional insight to the structure

of the layer. The data are presented in Figure 4.5.

A least square fit was applied to the Raman frequencies of the A1 phonon mode.

The fit function was a second order polynomial with a small square contribution. Thus

the dependence can be conceived as ’almost’ linear. The full width at half maximum

showed the same dependency: Peaks were narrow at the Cu-rich edge and linearly

broadening towards the Cu-poor edge.

It is known that absorber layers for effective solar cell devices must be grown under

4.2. Dependence on Stoichiometry 49

0.8 0.9 1.0 1.1 1.2 1.3 1.4292.0

292.5

293.0

293.5

294.0

294.5

295.0

295.5

Ra

ma

n S

hif

t (c

m-1

)

Cu-poor Cu-rich[Cu]/[In]

0.8 0.9 1.0 1.1 1.2 1.3 1.42

4

6

8

10

12

14

16

18

20

22


FW

HM

(c

m-1

)

Figure 4.5: (Left:) Frequency of the A1-phonon mode at 294 cm−1 of CuInS2 over the

composition the sample. A selection of the related spectra were given in

Figure 4.4. (Right:) Full width at half maximum of the A1-phonon mode.

Cu-excess conditions in the sequential process. Best results were obtained in the range

[Cu]/[In]=1.2 to 1.8 [75] where excess copper formed binary CuxS phases at the surface

of the films. The presence of the secondary copper sulfide was found to be crucial

for the crystal quality of the layers. Similar findings were reported for CuInS2 layers

prepared by coevaporation [76]. Klenk et. al [77] suggested a growth model for CuInSe2

prepared by coevaporation. Thereafter, a liquid layer of CuSe is present on top of the

growing CuInSe2. The vapor species condense at the surface of CuSe. They are then

transported to the binary/ternary interface where the crystallite is growing. The result

of this transport assistance are larger crystals with a better structural quality.

The Raman results can be explained in the framework of crystal and domain size. A

transmission electron microscope image from a CuInS2 layer prepared by coevaporation

is given in Figure 4.6.

Figure 4.6: TEM cross-section images of CuInS2 samples deposited from multisources

at 520 C (a) Cu rich and (b) Cu poor. Taken from [13].

The difference in crystal size can be easily recognized. Crystals from the Cu-rich

layer were up to 1 µm in size, whereas crystals from the Cu-poor layer are hardly


100 nm in diameter. Crystals size is defined by the grain boundaries. A crystal may be

subdivided in smaller units by stacking faults or twins. Those units without any lattice

distortions will be referred to as domains. They give rise to shifts towards lower Raman

frequencies and broadening of phonon modes when domains are smaller than 30 nm

[78]. The observed Raman shift of the A1-chalcopyrite mode was therefore attributed

to smaller domains in Cu-poor films.

The development of the A1-CuAu phonon frequency and the related FWHM over the

composition is given in Figure 4.7. There is a pronounced maximum and minimum

0.80 0.85 0.90 0.95 1.00 1.05303.0

303.5

304.0

304.5

305.0

305.5

306.0

306.5

307.0

307.5

308.0

308.5

309.0

Ram

an

Shift(cm

-1

)

Cu-richCu-poor [Cu]/[In]

0.80 0.85 0.90 0.95 1.00 1.05

10

12

14

16

18

20


FW

HM

(c

m-1

)

Figure 4.7: (Left:) Frequency of the A1-phonon mode of the CuAu-structure at 305

cm−1 over the composition of the sample. The spectra were given in Figure

4.4. (Right:) Full width at half maximum of the A1-phonon mode.

around [Cu]/[In]=0.93 in peak position and width, respectively. This findings were

interpreted as followed: The size of CuAu domains was maximal in the slightly copper

deficient region. This was expressed by the minimum in FWHM. The CuAu domains in

the stoichiometric region got smaller because the cations favor the chalcopyrite order.

When the composition was [Cu]/[In]<0.93 CuAu domains again got smaller because

not enough copper was available to build the structure.

Chapter 5

Reactive Annealing in H2S

A novel method for the preparation of thin CuInS2 films was introduced by this work.

It has been demonstrated that rapid thermal processing (RTP) can be employed for

the fabrication of high quality CuInS2 films by sulfurization of Cu-In precursors in

H2S. RTP systems offer precise control over process parameters such as temperature,

temperature ramp and pressure. CuInS2 film properties were studied and optimized in

dependence on those parameters.

Vibrational properties of absorber films depending on sulfurization temperature will

be presented first. It has been shown earlier that the existence of the CuAu-phase is

related to Cu-deficient films. Now, it will be demonstrated that the CuAu phase exists

also in CuInS2 films prepared from Cu-rich precursors at sulfurization temperatures

between 375 C and 500 C.

In order to analyze the region close to the Mo back contact by Raman spectroscopy

absorber layers were peeled off the substrate. It will be shown that the entire Cu-In

stack was sulfurized within a few seconds.

When CuInS2 films are prepared from Cu rich precursors the excess copper forms a

binary CuxS (0<x≤2) segregation at the surface. This segregation was analyzed in de-

pendence on sulfurization temperature by Raman spectroscopy and X-ray diffraction

measurements.

Thereafter the gathered information about the phase formation processes will be sum-

marized, including XRD precursor studies. The phase formation of CuInS2 in the

H2S-RTP process will be discussed on this basis.

The correlation of precursor morphology, sulfurization parameters and electronic prop-

erties of solar cell devices was studied. The presence of CuAu phase and its relation

to poor device performance will be discussed and an optimized sulfurization parameter

set will be presented. A further subject will be the influence of precursor morphol-

ogy on solar cell properties. It was found that the morphology of evaporated Cu-In

precursors is suitable for the preparation of homogeneous absorber films and efficient

devices. In contrast, the morphology of sputtered precursors lead to inhomogeneous

51

52 Chapter 5. Reactive Annealing in H2S

CuInS2 distribution and therefore to less efficient solar cells. It will be shown how this

difficulty can be overcome using the fast heat ramps of the RTP system.

5.1 Dependence on Sulfurization Temperature:

Phonon Modes for the Absorber Front Side

Phase formation of CuInS2 was studied by Raman spectroscopy. For the analysis a

series of samples was prepared from evaporated Cu-In stacks. The intention was the

conservation of the phases formed during heat up. Precursors were sulfurized in H2S

using the RTP system. The preset temperature was 550 C. The process was inter-

rupted during the heating up phase and the samples were quenched in a stream of

nitrogen. The temperature profiles are given in Figure 5.1. The temperature after

interruption is indicated by the dashed lines for three selected runs.

100

200

300

400

500

600

0

Processing time (min)

0 2 4 6 8 10 12 14

H S2

Tsulf

0 2 4 6 8 10

Sulfurization time (min)

Figure 5.1: Measured temperature profiles for a H2S sulfurization run(solid line) an

from interrupted runs (dashed lines). The interrupted runs were used to

quench the samples.

The Raman spectra for the samples are given in Figure 5.2. There is no Raman signal

for the sample quenched from 350 C. A double structure of peaks at 292 cm−1 and

308 cm−1 is present for samples annealed at temperatures between 375 C and 425 C.

The peaks correspond to the A1-phonon modes of the chalcopyrite and CuAu-ordered

lattice, respectively. There is no CuAu-phonon mode above 425 C. A phonon mode at

474 cm−1 for the secondary CuS-phase appears at 450 C. An increase in intensity of

the mode with increasing temperatures is obvious. The mode at 550 C is four times

5.1. Dependence on Sulfurization Temperature: Phonon Modes forthe Absorber Front Side 53

100 200 300 400 500

142

350 °C50

0

474308292

24162

550°C

525°C

500°C

475°C

450°C

425°C

400°C

375°C

In

tens

ity (

Cts

/mW

⋅ min

)

Raman Shift (cm-1)

Compound Obs. Lit. Ref.

(cm−1) (cm−1)

CuInS2-Ch 292 292+x This

work

241 244 [10]

CuInS2-CuAu 308 305+x This

work

CuS 474 475 [74]

142 142 [74]

62 62 [74]

Figure 5.2: Raman spectra for Cu-In metal stacks after sulfurization in H2S at various

temperatures. Temperature was ramped up to the values given above the

spectra and than the samples were cooled down in a nitrogen stream. The

spectra were recorded using the 514.5 nm line of an Ar+-laser. The phonon

modes and their assignments are compiled in the table.

more intense than the one for the sample at 450 C. Additionally, at 525 C and 550C it is accompanied by the less intense CuS modes at 62 cm−1 and 142 cm−1. The

increase in scattering intensity indicates a surface segregation of increasing thickness.

This could be caused by the higher temperatures themselves or the fact that it took

somewhat longer to reach higher temperatures (Figure 5.1) and thus annealing time

was increased. As reaction kinetic depend on both parameters CuS formation is prob-

ably promoted by both of them.

Peak positions and full widths at half maximum (FWHM) of the modes at 292 cm−1 and

305 cm−1 were evaluated and are depicted in Figure 5.3. The chalcopyrite A1-phonon

mode shifts towards lower Raman frequencies with increasing sulfurization tempera-

tures (Fig. 5.3a) and the related FWHM is decreasing (Fig. 5.3b). Above 450 C the

values remained almost constant at 292 cm−1 and 6.5 cm−1, respectively. The CuAu

A1-mode will be considered next. A narrow peak was found just after the appearance of


350 400 450 500 550291

292

293

294

295

296

297

Ch:CuInS2 - A

1 mode

(a)

Ram

an S

hift

(cm

-1)

Sulfurization Temperature (°C)

350 400 450 500 5505

6

7

8

9

10

11

12

13

14

15

16

17

Ch:CuInS2 - A

1 mode

(b)

FW

HM

(cm

-1)


350 400 450 500 550302

304

306

308

310

312

CuAu:CuInS2 - A

1 mode

(c)

Ram

an S

hift

(cm

-1)


350 400 450 5007

8

9

10

11

12

13

14

15

16

17

CuAu:CuInS2 - A

1 mode

(d)

FW

HM

(cm

-1)


Figure 5.3: (a) Raman shift and full width at half maximum (FWHM) (b) of the CuInS2

chalcopyrite A1-phonon mode. All values are plotted versus the sulfurization

temperature at which the sulfurization process was stopped. (c)Raman shift

and FWHM (d) of the CuInS2 CuAu-like ordered A1-phonon mode. The

lines are only a guide to the eye.

the mode at 375 C (Fig. 5.3d) and was broadening with higher temperatures. Raman

frequency of the CuAu A1-mode decreases with increasing temperatures (Fig. 5.3c)

and finally disappeared above 475 C.

Phonon modes from chalcopyrite and CuAu appear simultaneously at 375 C. Narrow

CuAu peaks indicate extended CuAu ordered domains. In comparison, chalcopyrite

ordered domains are smaller according to the broad A1-phonon mode.

Towards higher temperatures, the cations tend to the chalcopyrite order on expense of

the CuAu-order. This is expressed in the development of the peak width (FWHM) of

both phases: The shift of the chalcopyrite A1-phonon mode to lower Raman frequencies

is consistent with more extended domains.

5.2. Dependence on Sulfurization Temperature: Phonon Modes forthe Absorber Back Side 55

5.2 Dependence on Sulfurization Temperature:

Phonon Modes for the Absorber Back Side

Heat radiation reaches the precursor stack from the front side and H2S gas is

exclusively available at the surface during RTP-sulfurization. Therefore, the reaction

can not occur simultaneously throughout the film. There rather must be a growth

front. In order to find out about possible structural variations in comparison to the

front side, absorber films were removed from the Mo/Glas substrate and analyzed by

Raman spectroscopy. A metal sheet was glued onto the front side of the films and and

removed by a fast jerk. The remaining shiny Mo proved the complete release of the

films. Samples from the quenching series (Section 5.1) were used for the experiments.

The recorded spectra are shown in Figure 5.4.

100 200 300 400 500

550 °C

500 °C

425 °C

375 °C

474

410

386

352

325

305

294

265

245

350

Inte

nsity

(C

ts/m

W⋅m

in)

Raman Shift (cm-1)

Compound Obs. Lit. Ref.(cm−1) (cm−1)

CuInS2-Ch 245 244 [10]265 266 [10]241 244 [10]294 294 This

work325 321 [10]352 352 [10]

CuInS2-CuAu 308 305 Thiswork

CuS 474 475 [74]MoS2 386 382 [79]

410 407 [79]

Figure 5.4: (Left:) Raman spectra from the back side of CuInS2 films close to the Mo

contact. The back side was made accessible by peeling off the films from the

substrate. Samples were taken from the H2S temperature series described in

Section 5.4. Sulfurization temperature was ramped up to the given values.

Thereafter the samples were cooled down in a nitrogen stream. The spectra

were recorded using the 514.5 nm line of an Ar+-laser. The penetration

depth for this wavelength is about 100 nm. (Right:) Assignment of the

observed phonon modes.


The spectra for films sulfurized at temperatures between 375 C and 500 C, show

exclusively CuInS2 chalcopyrite and CuAu phonon modes. The spectra for the film

sulfurized at 550 C showed additionally MoS2 phonon modes at 386 cm−1, 410 cm−1

and a CuS mode at 474 cm−1. As carried out throughout this work, spectra were

recorded for five different spots on the same sample (refer to Section 3.2). MoS2

phonon modes were observed for all the measured spots on the sample. In contrast,

CuS modes were just found for one spot. Apparently, MoS2 was present in extended

areas, whereas CuS was present in form of isolated islands.

CuxS (0<x≤2) segregation at the front surface was removed by KCN etching for the

preparation of solar cells. It is assumed that after etching no CuxS remains neither

on the surface nor in the bulk. This assumption is justified by the negative results

of XRD and reasonable solar cell performance. CuS is a highly conductive material

and would give rise to shunt paths if present. Therefore it is surprising to detect

CuS at the absorber back side. On the other hand side, it will be shown later that

sulfurization temperatures above 525 C cause decrease in conversion efficiency for

solar cells. Thus it can be assumed that such an decrease in efficiency is related to the

presence of CuS at the absorber back side.

Next, the chalcopyrite and CuAu A1-phonon mode will be considered. Raman

frequencies and FWHM of the modes are given in Figure 5.5. Phonon mode for

CuInS2 structures were observed for Tsulf ≥ 375 C. This temperature is also the

critical temperature for the existence of CuInS2 phonon modes at the front side of the

film (compare with Figure 5.2). The exposition of the precursors to the maximum

temperature lasted just a few seconds. Therefore it was concluded, that the initial

formation of chalcopyrite/CuAu occurred within this period.

Raman shift and FWHM of the chalcopyrite A1-mode versus the temperature show

the same trends as those for the front side. Thus chalcopyrite domains seem to grow

almost simultaneously at the front and back side. Phonon frequency and peak width

of the CuAu A1-phonon show also the same trend as for the front side but there is a

difference in the dependence on the sulfurization temperature. Frequency and FWHM

for the back side in Figure 5.5c+d decrease and increase, respectively, more steep at

lower sulfurization temperatures than at the front side. This indicates that extended

CuAu domains were conserved up to higher temperatures at the front side than at the

back side.

The formation of CuAu ordered domains has been associated with films prepared

from Cu-poor precursors earlier in this work. Here, CuAu domains formed also from

Cu-rich precursors annealed in H2S. In both cases, CuAu domains form simultaneously

with chalcopyrite domains. It is obvious that the presence of CuAu domains prevents

the formation of single phase chalcopyrite grains necessary for efficient solar cell

devices. Two preparation rules can be deduced from this findings: First, the metal

ratio for the precursor stack should be at least [Cu]/[In] ≥ 1.1. This is in agreement

5.2. Dependence on Sulfurization Temperature: Phonon Modes forthe Absorber Back Side 57

350 400 450 500 550293

294

295

296

297

298

(a)

Ram

an S

hift

(cm

-1)

Ch:CuInS2 - A

1 mode

Sulfurization Temperature (cm-1)

350 400 450 500 5504

5

6

7

8

9

10

11

(b)

FW

HM

(cm

-1)

Ch:CuInS2 - A

1 mode


350 400 450 500 550304

305

306

307

308

309

310

311

312

(c)

Ram

an S

hift

(cm

-1)

CuAu:CuInS2 - A

1 mode


350 400 450 500 55010

11

12

13

14

15

16

17

(d)

FW

HM

(cm

-1)

CuAu:CuInS2 - A

1 mode


Figure 5.5: (a) Raman shift and full width at half maximum (FWHM) (b) of the CuInS2

chalcopyrite A1-phonon mode. All values are plotted over the sulfurization

temperature where the sulfurization process was stopped. (c)Raman shift

and FWHM (d) of the CuInS2 CuAu-like ordered A1-phonon mode. The

solid lines are only a guide to the eye.

with reports from other authors. Scheer et. al [37] prepared device grade absorber

films by coevaporation with [Cu]/[In] = 1.0 - 1.8. Gossla et al. [80] obtained quality

films by H2S sulfurization of Cu-In stacks with [Cu]/[In] = 1.3. Second, the optimal

sulfurization temperature for Cu-rich precursors is Tsulf = 525 C. This temperature

is just high enough to obtain single phase chalcopyrite films and it is low enough to

prevent CuS remainder at the back side.


5.3 Surface Segregation

After sulfurization of Cu-rich precursors in elemental sulfur CuS (covellite) is present

at the surface of the CuInS2 films [38, 75]. The CuS segregation can be easily identified

by optical inspection due to its typical indigo-blue color. In contrast, the surfaces of

CuInS2 films prepared by H2S sulfurization of Cu-In precursors with a high Cu-excess

([Cu]/[In] = 1.2) were always light grey and did not show the indigo-blue color of

CuS. In this section it will be shown that a surface segregation exists also on absorber

films prepared by H2S sulfurization. Furthermore it will be shown that the segregation

consists mainly of Cu9S5 digenite and small amounts of CuS.

Scanning electron microscopy (SEM) images of absorber layers before and after etching

are shown in Figure 5.6. Sharp grain edges and pronounced distinctions in height can

Figure 5.6: SEM images (Left:) from an as grown absorber layer prepared by H2S

sulfurization at 525 C for 5 min. The blurred structure was attributed to

a CuxS surface coating which could be removed by etching. (Right:) After

the KCN etching procedure. The coating disappeared.

be recognized on the image from the etched sample. In contrast, a coating on the

as grown sample surface is indicated by the blurred structures. The coating can be

attributed to the CuxS surface segregation.

The segregation was analyzed with Raman spectroscopy and X-ray diffraction. Raman

spectra from samples prepared from Cu-rich precursors ([Cu]/[In] = 1.2) and sulfurized

under standard conditions (Tsulf = 525 C, tsulf = 5 min) are given in Figure 5.7. The

spectra for the ”as grown” samples show strong Raman lines at 62 cm−1 and 474

cm−1 besides the chalcopyrite phonon modes. After etching the samples in KCN those

lines disappeared completely (lower spectrum). A CuS single crystal was measured for

reference. Its phonon modes matches the modes quite well. In conclusion, CuS has

5.3. Surface Segregation 59

100 200 300 400 500

10

00

CuInS2 film

KCN etched

CuInS2 film

as grown

CuSsingle crystal

62

142 267

474

Inte

nsity

(ct

s /

mW

⋅ min

)

Raman Shift (cm-1)

Figure 5.7: Raman spectra prior and

after KCN etching of an

CuInS2 film. For compar-

ison the spectrum of a CuS

single crystal is given.

been proven to occur at the surface of the absorber film. Furthermore its complete

removal by KCN etching has been demonstrated.

X-ray diffraction in grazing incidence was performed to determine the presence of

possible CuxS phases on the surface of absorber layers. But no peaks besides the CuInS2

reflections were observed. There is still the possibility that reflections from CuxS phases

coincide with those from the chalcopyrite structure. To check out this possibility CuxS

binaries were synthesized: Thin layers of Cu were deposited and sulfurized in H2S at

varying temperatures. The sulfurization time was 5 minutes. XRD-GI spectra for the

sulfurized Cu films are shown in Figure 5.8 (on the left).

The spectra for the Cu film sulfurized at 475 C matches the data for Cu31S16

djurleite [81], a phase next to Cu2S in the phase diagram (Appendix A). Reflections

for Cu films sulfurized at 550 C, 525 C and 500 C were assigned to Cu9S5 digenite.

Comparison of the digenite spectra with the chalcopyrite reflections shown in Figure 5.8

reveals that digenite reflections coincide with those for chalcopyrite. Reviewing all the

entries for Cu-S binaries in the JCPDS database [81] yields that digenite reflections

are the only ones completely coinciding with chalcopyrite reflections. This strongly

supports the assumption that Cu9S5 is present on the absorber surface.

CuS was not detected by XRD-GI even though the Cu films sulfurized at 525 C and

550 C were slightly blue, indicating the presence of CuS. In contrast, the Cu films

sulfurized at 475 C and 500 C were grey. In order to verify the possible presence of

CuS Raman spectra were recorded. The spectra are shown in Figure 5.9.

No Raman signal is present for Cu films sulfurized at 475 C and 500 C. This

suggests that djurleite and digenite are not Raman active with respect to the excitation

wavelength (514.5 nm). However, Raman measurements for the Cu-S systems are only

reported for CuS [74] and Cu2S [82] and no literature data are available for CuxS


20 25 30 35 40 45 50 55 60

2 Theta

(28

2)

(84

2)

(84

0)

(01

5)

(35

0)

(02

3)

(72

1)

(52

3)

(37

3)

(10

1)

(00

15)

(10

7)

(10

10)

Mo

(11

0)

(11

0)

(11

15)

100

XR

DIn

tensity

(cts

/sec)

475 °C

500 °C

525 °C

550 °C

Cu

S:

Dju

rleite

31

16

Cu

S:D

jurleite

95

2

(10

7)

20 25 30 35 40 45 50 55 60

Theta

(00

15)

(10

10)

Mo

(11

0)

(11

0)

(11

15)

(11

2)

(10

3) (2

04)

(31

2)

XR

DIn

tensity

(cts

/sec)

20

0

CuIn

S:

Chalc

opyrite

2C

uS

:D

igenid

e9

5

Figure 5.8: XRD-GI spectra (glance angle 2.5) (Left:) for Cu films sulfurized in H2S

at varying temperatures (tsulf = 5 min). (Right:) Spectrum for the 525 C

sample together with the spectrum for a CuInS2 film sulfurized at Tsulf =

525 C, tsulf = 5 min.

(1<x<2) phases for comparison.

Phonon modes from CuS were observed for films sulfurized in H2S at 525 C and 550C. Raman measurements are more surface sensitive than XRD measurements. The

negative result from the XRD measurements suggests therefore a comparable small

contribution of CuS to the digenite surface segregation.

There is no direct of proof for Cu9S5 digenite to occur at the CuInS2 surface. But in

view of the presented results it is concluded that digenite yields the main contribution

to the surface segregation. Additionally small amounts of CuS are formed.

5.4. Phase Formation by H2S 61

100 200 300 400 500

475 °C

142

550 °C

525 °C

500 °C250

62

267

474

Inte

nsity

(ct

s /

mW

⋅ min

)

Raman Shift (cm-1)

Figure 5.9: Raman spectra for thin

Cu films sulfurized at vary-

ing temperatures for 5 min.

Excitation wavelength was

λ = 514.5 nm.

5.4 Phase Formation by H2S

Results from the previous sections concerning the phase formation of CuInS2 are sum-

marized in Figure 5.10. Furthermore, new information from XRD measurements were

added. They refer to the Cu-In precursors sulfurized in the quenching series (Section

5.1).

CuIn2 and free Cu were found a day after the precursors ([Cu]/[In]=1.2) were deposited

at room temperature. XRD reflections for the In-rich CuIn2 was detected up to 300C. At higher temperatures the alloy changed to the Cu-rich Cu16In9 phase which was

found up to 425 C.

CuIn2 forms below room temperature and is stable up to 148 C according to the

report of Keppner et al. [41]. The absence of In reflections in the XRD spectra taken

from the metal stack a day after evaporation indicated the complete alloying of the

metals. In the quenching experiments CuIn2 was observed up to 300 C which was

attributed to the fast heating up in the RTP system. Cu16In9 is the only binary phase

which was formed during the sulfurization process. It was detected up to 425 C.

Phonon modes from chalcopyrite and CuAu structures were detected in a early stage

of the sulfurization, Tsulf = 375 C. From the binary compounds solely Cu16In9 was

detected at this temperature. Thus it can be concluded that CuInS2 forms directly

from the Cu16In9 alloy. In Section 5.1 it was shown that CuAu order is preferred

over the chalcopyrite order at this temperature. The cations rearrange with higher

temperatures in favor of the chalcopyrite order. CuAu domains disappear above 475C and CuInS2 material remains as the only ternary compound. At the same stage

CuS appears. CuS is not stable at temperatures above 507 C according to the phase

diagram. In Section 5.3 it was shown that Cu9S5 can be found above 500 C. So it is

likely that CuS turns into Cu9S5 above 500 C. Raman measurements revealed that


Cu

CuS

Cu S9 5

CuInS : CuAu2

CuInS : Ch2

Cu In16 9

CuIn2

0 15 30 60 75 90 105 120 135 150 165 180 450

~~~~

30

0°C

40

0°C

42

5°C

45

0°C

47

5°C

50

0°C

20

°C

55

0°C

435

Processing period (sec)

Sulfurization temperature

from XRD

from Raman

from Raman and XRD

52

5°C

55

0°C

~~~~

~~

Figure 5.10: Phase formation sequence of CuInS2 absorber films from Cu-In stacks

([Cu]/[In] = 1.2) sulfurized in H2S. Heat ramp: ∆T/∆t = 3C / sec.

CuS is present even after 5 minutes of sulfurization at Tsulf = 550 C. Thus the phase

transition was not completed and CuS partly remains.

5.5 Photovoltaic Performance: Dependence on

Morphology and Structure

Cu-In precursor films can be deposited by sequential sputter and evaporation tech-

niques. The resulting precursor morphology and homogeneity will be discussed with

respect to the photovoltaic performance of CuInS2 solar cells. The influence of H2S

sulfurization parameters on the absorber layers was studied by analysis of the electronic

properties of the related solar cell devices.

Cu-In precursors were prepared by sequential sputtering and evaporation for a metal

ratio [Cu]/[In] = 1.2. Scanning electron microscopy (SEM) images for a sputtered and

an evaporated precursor are shown in Figure 5.11(a) and (c), respectively.

In case of the sputtered precursors an almost closed top layer can be observed.

After sulfurization in H2S (Tsulf = 525 C, ∆T/∆t = 3 C/sec , tsulf = 5 min)

protuberant aggregations of crystals were observed on the surface by SEM. One of

them is shown in Figure 5.11(b). CuInS2 is the only phase proven by XRD, therefore

the crystal aggregations are CuInS2 as well. They are isolated from each other and

103 of them were found per cm2. Their formation can be avoided by setting the heat

ramp to ∆T/∆t ≥ 9 C/sec.

In contrast, the sequential evaporation of Cu-In leads to the formation of distinct

5.5. Photovoltaic Performance: Dependence on Morphology andStructure 63

(a) 4µm 10µm(b)

CuInS2 1µm(d)4µm(c)

Figure 5.11: SEM image of a sequentially sputtered Cu/In precursor (a). Stoichiomet-

ric CuInS2 segregation formed on a Cu-rich CuxS/CuInS2 absorber layer

(b). Evaporated Cu/In precursor (c). Cross section of a CuInS2 layer

prepared from an evaporated Cu/In precursor by H2S sulfurization (d).

islands as shown in Figure 5.11(c). No protuberant aggregations are observed after

sulfurization of this precursors. A cross section is shown in Figure 5.11(d). No

significant differences in average grain size can be found for sulfurization temperatures

in the range from 475 C to 550 C. The structure of sequentially evaporated and

sputtered precursors right after deposition are sketched in Figure 5.12(1) and (I),

respectively. It is known that copper and indium form CuIn2 at room temperature.

It was shown in Section 5.4 that a day after deposition the entire In of such Cu-In

bilayers reacts to CuIn2 and only free Cu remains. Therefore the structures observed

in Figure 5.11(c) can be attributed to CuIn2 on copper as the images were taken the

day after preparation. It is assumed that the morphology of the initial metal layers

was conserved while alloying (Figure 5.12(2)).

It seems that In of sequentially sputtered precursors was more homogeneous dis-

tributed after deposition (Figure 5.12(II)) than In of evaporated films which tends to

form isolated islands.


CuIn

Cu

CuIn2

Mo

MoMo

Mo

Cu

CuIn2

Mo

CuInS2

Cu S + CuS9 5

5 % H S in Ar2

Cu Sx

CuIn

Mo

CuCuIn

2

Mo

CuCuIn

2

Mo

5 % H S in Ar2

CuIn2

Mo

CuInS2

Cu S + CuS9 5

CuInS2

0 h after deposition:

10 h after deposition:

Annealing in H S:At T(10 sec) ~ 100 °C

2

~

After sulfurization:Resulting structure

Evaporated: Sputtered:

(1) (I)

(2) (II)

(3) (III)

(4) (IV)

Figure 5.12: Formation of CuInS2 films from sequentially evaporated and sputtered

precursors ([Cu]/[In] = 1.2), respectively. (1:) Dewetting of the In film

after deposition. (2,II:) Precursor alloying, conservation of the morphol-

ogy. (3,III:) Snap-shot of the sulfurization process. (4,IV:) Film structure

after sulfurization.

5.5. Photovoltaic Performance: Dependence on Morphology andStructure 65

For the explanation of the observed protuberant aggregations it is assumed that

CuIn2 pellets are formed on the surface (Figure 5.12(III)) during heating up in the RTP

system. Due to the local accumulation of the alloy crystal aggregations are formed. In

contrast, no aggregations were observed after sulfurization of evaporated precursors.

Obviously, the island structure hinders the accumulation of CuIn2. A homogeneous

CuInS2 layer was obtained. The copper film, accessible through the gaps between the

CuInS2 islands, forms probably a CuxS phase at the surface in the initial stage of H2S

sulfurization (Figure 5.12(3)). The presence of this CuxS phase is may be responsible

for the suppression of the dewetting effect. In result a more homogeneous CuInS2 film

is obtained.

Absorber morphology and homogeneity is reflected in the performance of the solar cell

devices. Photovoltaic parameters of CuInS2 based devices from sputtered precursors

versus heat ramp ∆T/∆t are shown in Figure 5.13. The sulfurization temperature was

525 C.

2 4 6 8 10 12 14 16 18 20 22450

500

550

600

650

700

750

Tsulf

= 525 °C t

sulf = 5 min

VOC

Cur

rent

den

sity

(m

A/c

m2 )

Heat ramp: ∆T/∆ t (°C/sec)

Ope

n ci

rcui

t vo

ltage

(m

V)

6

8

10

12

14

16

18

20

ISC

0

2

4

6

8

10

Agg.

Agg

rega

tion

dens

ity (

x103 c

m-2)

2 4 6 8 10 12 14 16 18 20 22

50

52

54

56

58

60

62

64

66

68

70

Tsulf

= 525 °C t

sulf = 5 min

ff

Con

vers

ion

effi

cien

cy (

%)

Heat ramp: ∆T/∆ t (°C/sec)

Fill

fac

tor

(%)

2

4

6

8

10

η

0

2

4

6

8

10

Agg.

Ag

gre

ga

tion

de

nsi

ty (

x103 c

m-2)

Figure 5.13: Photovoltaic parameters of CuInS2 solar cells prepared by H2S sulfuriza-

tion of sputtered precursors using different heat ramps. For comparison,

density of aggregations like in Figure 5.11(b) have been added. The lines

are only a guide to the eye.

For comparison, the density of the surface aggregations were added. Conversion

efficiencies η > 8 % can only be obtained when the temperature is ramped up at

∆T/∆t ≥ 9 C/s. Below this threshold a drastic decrease in efficiency is observed.

The simultaneous decrease of all parameters indicate shunt paths which are most likely

caused by the inhomogeneous lateral CuInS2 distribution.

In the case of evaporated Cu-In layers no dependence on the heat ramp was observed.

Heat ramp could therefore be set as low as ∆T/∆t = 3 C/s in order to avoid possible

thermal stress to the glass substrate. Photovoltaic parameters were determined in

dependence on H2S sulfurization temperature and are shown in Figure 5.14.


400 425 450 475 500 525 550

0

100

200

300

400

500

600

700

800

Cur

rent

den

sity

(m

A/c

m2 )

Sulfurization temperature (°C)

Ope

n ci

rcui

t vo

ltage

(m

V)

0

4

8

12

16

20

24

6

8

10

12

14

CuAu: FWHM

Ch: FWHM

ISC

VOC

FW

HM

(cm

-1)

400 425 450 475 500 525 550

0

10

20

30

40

50

60

70

Con

vers

ion

effi

cien

y (%

)

Sulfurization temperature (°C)F

ill f

acto

r (%

)

0

2

4

6

8

10

12

6

8

10

12

14

CuAu: FWHM

Ch: FWHM

η

ff

FW

HM

(cm

-1)

Figure 5.14: Photovoltaic parameters of CuInS2 solar cells prepared from evaporated

Cu-In precursors. The H2S sulfurization temperature was varied. Heat

ramp was ∆T/∆t = 3 C/s and tsulf = 5 min. FWHM data for the chal-

copyrite and CuAu A1 phonon modes were added for comparison. The

lines are only a guide to the eye.

The highest conversion efficiency is 11 % at Tsulf = 525 C. Above 525 C there is

a significant decrease in efficiency due to a loss in current and fill factor. Earlier it was

shown that CuS is present close to the back contact for Tsulf = 550 C. It is possible

that the loss is caused by CuS due to its high conductivity. Between 525 C and 475C efficiencies > 8 % were achieved. Thus there exists a broad temperature window for

the H2S sulfurization process. For Tsulf < 475 C a strong decrease in all parameters

is observed and at 425 C there is no energy conversion. Although CuInS2 is formed

already at 375 C, the electronic film character is metallic due to Cu-In alloys present

up to 425 C (refer to Figure 5.10).

FWHM data for the chalcopyrite and CuAu A1 phonon modes were added for compar-

ison to Figure 5.14. The increase in efficiency between 425 C and 475 C correlates

with the disappearance of CuAu domains in the absorber films. High efficiencies were

only obtained when phonon modes for the CuAu structure completely disappeared.

It can be summarized that the suitability of the described H2S-RTP process for the

preparation of efficient CuInS2 solar cells depends on the structure of the used pre-

cursors. The inhomogeneous distribution of CuInS2 crystals in case of the sputtered

precursors can only be avoided by using heat ramps ∆T/∆t ≥ 9 C/s. This problem

does not occur for evaporated precursors. For this reason investigations in this work

focused on absorber films prepared from sequentially evaporated precursors.

The photovoltaic device performance is correlated to the size of CuAu present in the

absorber films. CuAu domains must be avoided in order to obtain efficient solar cells.

Chapter 6

Summary

Thin CuInS2 films are used as absorber material for solar cells. In the framework of

this thesis films were prepared by reactive annealing of Cu-In precursors in H2S gas.

A rapid thermal process was employed for it the first time. Raman spectroscopy was

used for the analysis of the CuInS2 films.

The vibrational properties of CuInS2 films depend on the [Cu]/[In] ratio offered for the

chemical reaction. Apart from the known phonon modes of the chalcopyrite lattice,

additional phonon modes appeared for [Cu]/[In] < 1 at 305 cm−1 and 60 cm−1. This

phonon modes can also be observed for films prepared by reactive annealing of Cu-rich

precursors in the temperature range 375 C - 475 C. The phonon mode at 305 cm−1 is

A1-symmetric as demonstrated by polarization depend Raman measurements. A poly-

morphous structure of the chalcopyrite lattice, the so called CuAu order, is responsible

for the additional phonon modes. This was proven by group theoretical analysis and

phonon frequency calculations.

Reaction kinetics was investigated by means of combined Raman and X-ray diffraction

measurements. Chalcopyrite and CuAu ordered structures form in the early stage of

heating up from the precursor. CuAu order is preferred in this stage. The fraction of

CuAu ordered domains decreases with increasing temperature and is finally vanishing.

The excess copper forms Cu9S5 and CuS on the surface. The Cu-S binaries can be

completely removed by chemical etching.

CuInS2 solar cells based on the prepared CuInS2 films were characterized in depen-

dence on the preparation parameters. It was shown that loss in conversion efficiency

is related to the appearance of CuAu-ordered domains. Single phase CuInS2 films of

high crystalline quality could be prepared by reactive annealing of evaporated Cu-In

films in H2S for 5 minutes at 525 C. Solar cells with efficiencies up to 11 % could be

procuced on basis of this films. The used Raman setup together with the data from

this work can be utilized for ex-situ quality control of CuInS2 absorber films. Moreover,

the fundamentals for the development of a Raman setup for in-situ growth control are

provided.

67

Appendix A

CuS Phase Diagram

Figure A.1: Cu-In phase diagram [86].

68

69

Table A.1: Phases of the Cu-S system [86].

Composition Pearson Space ProtoPhase at.% S(Cu/S) symbol group type(Cu) 0 cF4 Fm3m Cuα chalcocite (αCu2S) 33.33 mP144(?) P21/cβ chalcocite (βCu2S) 33.3 hP6 P63/mmc InNi2Djurleite(Cu 1.96S) 33.7 to 34.1(a) oP380(?) Pmnm

P21nm(?)Pmn21

Digenite (Cu2−δS) 35.5 to 36.2(b) cF12 Fm3m CaF2

Anilite (Cu1.75S) 36.36±0.04 oP44(?) PnmaCovellite (CuS) 50 hP12 P63/mmc CuS(S) 100 oF128 Fddd αS

mP48 P21/α βShR6 R3 εS

Metastable phasesProtodjurleite 33.7 (1.97)(c)

33.8 (1.96)(d)Tetragonal 33.8 (1.96) P43212 Ge III(HP)Hexagonal-tetragonal CuxS 34.1 to 36.4 (1.93 to 1.75)Low digenite (αDg) 35.84 to 36.15 (1.790 to 1.766)(e) R3mBlaubleibender covellite I 41.7±1.7 (1.4±0.1)Blaubleibender covellite II 47.7±2.3 (1.1±0.1)CuS2 66.67 (0.5) Pa3(?)

(a) At 72 C. (b) At 80 C. (c) At 75 C. (d) At 93 C. (e) At 25 C.

Bibliography

[1] Shell AG. Energie im 21. Jahrhundert. Aktuelle Wirtschaftsanalysen, 25, 1995.

[2] R.W. Birkmire. Compound Polycrystalline Solar Cells: Recent Progress and Y2K

Perspecitve. Solar Energy Materials and Solar Cells, 65:17–28, 2001.

[3] M.A. Contreras, B. Egaasa, K. Ramanathan, J. Hiltner, A. Swartzlander, F. Ha-

soon, and R. Noufi. Progress Towards 20 % Efficiency in Cu(In,Ga)Se2 Polycrys-

talline Thin-film Solar Cells. Progr. Photovolt.: Res. Appl., 7(4):311–316, 1999.

[4] R. Klenk, P. Dobson, M. Falz, N. Janke, I. Luck, A. Perez-Rodrıguez, R. Scheer,

and E. Terzini. Sulfurcell: Efficient Thin Film Solar Cells Based on CuInS2. Proc.

of 16th EPVSEC, 2000.

[5] K. Siemer, J. Klaer, I. Luck, J. Bruns, R. Klenk, and D. Braunig. Efficient CuInS2

solar cells from a rapid thermal process (RTP). Solar Energy Materials and Solar

Cells, 67:159–166, 2001.

[6] F. Karg, V. Probst, H. Harms, J. Rimmasch, W. Riedl, J. Kotschy, J. Holz, R. Tre-

ichler, O. Eibl, A. Mitwalsky, and A. Kiendl. Novel Rapid-Thermal-Processing for

CIS Thin-Film Solar Cells. Proc. of 23th IEEE PVSEC, pages 441–446, 1993.

[7] G.D. Mooney, A.M. Hermann, J.R. Tuttle, D.S. Albin, and R. Noufi. Forma-

tion of CuInSe2 thin films by rapid thermal recrystallization. Appl. Phys. Lett.,

58(23):2678–80, 1991.

[8] F.W. Ohrendorf and H. Haeuseler. Lattice Dynamics of Chalcopyrite Type Com-

pounds. Part I. Vibrational Frequencies. Cryst. Res. Technol., 34(3):339–349,

1999.

[9] W. H. Koschel, F. Sorger, and J. Baars. Optical Phonons in I-III-VI2 Compounds.

J. Physique, C3:177–181, 1975.

[10] W. H. Koschel and M. Bettini. Zone-Centered Phonons in AIBIIIS2 Chalcopyrites.

Phys. stat. sol (b), 72:729–737, 1975.

70

Bibliography 71

[11] R. Bacewicz, W. Gebicki, and J. Filipowicz. Raman Scattering in CuInS2xSe2(1−x)

mixed crystals. J. Phys.: Condens. Matter, 6:L777–780, 1994.

[12] K.I. Kondo, S. Nakamura, and K. Sato. Raman Scattering Studies of CuInS2

Films Grown by RF Ion Plating. Jpn. J. Appl. Phys., 37(10), 1998.

[13] J. Alvarez-Garcıa, A. Perez-Rodrıguez, A. Romano-Rodrıguez, and J.R. Morante.

Microstructure and secondary phases in coevaporated CuInS2 films: Dependence

on growth temperature and chemical composition. J. Vac. Sci. Technol. A, 19(1),

2000.

[14] Shay and Wernick. Ternary Chalcopyrite Semiconductors: Growth, Electronic

Properties, and Applications. Pergamon Press, 1975.

[15] S.H. Wei, S.B. Zhang, and A. Zunger. Band structure and stability of zinc-blende-

based semiconductor polytypes. Phys. Rev. B, 59:R2478–R2481, 1999.

[16] J.E. Jaffe and Alex Zunger. Electronic structure of the ternary chalcopyrite semi-

conductors CuAlS2, CuGaS2, CuInS2, CuAlSe2, CuGaSe2 and CuInSe2. Phys.

Rev. B, 28(10):5822–5847, 1983.

[17] W. Borchardt-Ott. Kristallographie. Springer, 1997.

[18] S.H. Wei, L.G. Ferreira, and A. Zunger. First-principles calculation of the order-

disorder transition in chalcopyrite semiconductors. Phys. Rev. B, 45(5):2533–2536,

1992.

[19] D.S. Su and S.H. Wei. Transmission electron microscopy investigation and first-

principles calculation of the phase stability in epitaxial CuInS2 and CuGaSe2 films.

Appl. Phys. Lett., 74:2483–2485, 1999.

[20] T.Hahn. International Tables for Crystallography, volume A Space-Group sym-

metry. Kluwer Academic Publisher, 1995.

[21] M.I. Alonso, K. Wakita, J. Pascual, M. Garriga, and N. Yamamoto. Optical

functions and electronic structur of CuInSe2, CuGaSe2, CuInS2, and CuGaS2.

Phys. Rev. B, 63(075203):1–13, 1984.

[22] J.E. Jaffe and Alex Zunger. Anion displacements and the band-gap anomaly in

ternary ABC2 chalcopyrite semiconductors. Phys. Rev. B, 27(8):5176–5179, 1983.

[23] J.L. Shay. Experimental Studies of Energy Band Structure of Chalcopyrite Crys-

tals. Journal de Physique, Colloque C3, C3(9):C3–109–113, 1975.

[24] J.L. Shay, B. Tell, H.M. Kasper, and L.M. Schiavone. p-d Hybridization of the

Valence Bands of I-III-VI2 Compounds. Phys. Rev. B, 5(12):5003–5005, 1972.

72 Bibliography

[25] P. Y. Yu and Manuel Cardona. Fundamentals of Semiconductors. Springer, 1996.

[26] N.N. Syrbu, R.V. Cretu, and V.E. Tezlevan. Energy Band Structure of CuInS2

Crystals. Cryst. Res. Technol., 1:135–144, 1998.

[27] H. Ibach and H. Luth. Festkorperphysik. Springer, 1995.

[28] G.D. Boyd, H. Kasper, and J.H. McFee. Linear and nonlinear optical properties

of AgGaS2, CuGaS2, and CuInS2, and theory of the wedge technique for the

measurement of nonlinear coefficients. IEEE J. Quant. Electron., QE-7(12):563–

73, 1972.

[29] D.C. Look and J.C. Manthuruthil. Electron and hole conductivity in CuInS2. J.

Phys. Chem. Solids, 37(2):173–80, 1976.

[30] B. Tell and J.L. Shay. Electrical Properties, Optical Properties, and Band Struc-

ture of CuGaS2 and CuInS2. Phys. Rev. B, 4(8):2463–2471, 1971.

[31] K. Topper. Photolumineszenz an polykristallinen CuInS2-Absorberschichten und

Solarzellen. Master’s thesis, TU Berlin, 1996.

[32] I.V. Bodnar, A.G. Karoza, and G.F. Smirnova. Lattice reflectivity spectra of

CuGaS2-CuGaSe2 solid solutions. Phys. Stat. Sol. B, 84:K65–69, 1977.

[33] C. Carlone, D. Olego, Y. Jayaraman, and M. Cardona. Pressure dependence of

the Raman modes and pressure-induced phase changes in CuGaS2 and AgGaS2.

Phys. Rev. B, 22:3877–85, 1980.

[34] A. Anedda, G. Bongiovanni, and F. Raga. Raman spectra of CuGaS2-xSex (x=0,

0.5, 1.0). Nuovo Cimento D, 2:1950–6, 1983.

[35] J. Gonzalez and B.J. Fernandez. High-pressure behavior of Raman modes in

CuGaS2. Phys. Rev. B, 46:15092–101, 1992.

[36] J. Gonzalez and J. Guinet. Temperature dependence of first-order Raman scat-

tering in CuGaS2. Jpn. J. Appl. Phys. Suppl., 32-3:578–80, 1993.

[37] R. Scheer, T. Walter, H.W. Schock, M.L. Fearheiley, and H.J. Lewerenz. CuInS2

based thin solar cell with 10.2 % efficiency. Appl. Phys. Lett., 63(24):3294–96,

1993.

[38] R. Klenk, U. Blieske, V. Dieterle, K. Ellmer, S. Fiechter, I. Hengel, A. Jager-

Waldau, T. Kampschulte, Ch. Kaufmann, J. Klaer, M. Ch. Lux-Steiner,

D. Braunger, D. Hariskos, M. Ruckh, and H.W. Schock. Properties of CuInS2

thin films grown by a two-step process without H2S. Solar Energy Materials and

Solar Cells, 49:349–356, 1997.

Bibliography 73

[39] T. Watanabe and M. Matsui. Solar Cells Based on CuInS2 Thin Films through

Sulfurization of Precursors Prepared by Reactive Sputtering with H2S Gas . Jpn.

J. Appl. Phys., 35(3):1681–84, 1996.

[40] J.J.M. Binsma, L.J. Giling, and J. Bloem. Phase Relations In The System Cu2S-

In2S3. J. Cryst. Growth, 50:429–436, 1980.

[41] W. Keppner, T. Klas, W. Korner, R. Weschke, and G. Schatz. Compound For-

mation at Cu-In Thin-Film Interfaces Detected by Perturbed γ − γ Angular Cor-

relations. Phys. Rev. Letters, 54(21):2371–74, 1985.

[42] H. Okamoto. Cu-In Phase Diagram. Journal of Phase Equilibria, 15(2), 1994.

[43] M. Gossla, H. Metzner, and H. E. Mahnke. Coevaporated Cu-In films as precursors

for solar cells. J. Appl. Phys., 86(7):3624–32, 1999.

[44] H. Metzner, M. Brussler, K.-D. Husemann, and H. J. Lewerenz. Characterization

of phases and determination of phase relations in the Cu-In-S system by γ − γ

perturbed angular correlations. Phys. Rev. B, 44(21):11614–23, 1991.

[45] M. Gossla, H.-E. Mahnke, and H. Metzner. Investigation of thin films of the Cu-In

and CuInS2 system. Thin Solid Films, 361-362:56–60, 2000.

[46] C. Dzionk, H. Metzner, H. J. Lewerenz, and H.-E. Mahnke. Perturbed angular

correlations study of thin Cu-In films. J. Appl. Phys., 78(4):2392–97, 1995.

[47] A.E. Rakhshani and A.S. Al-Azab. Characterization of CdS films prepared by

chemical-bath deposition. J. Phys.: Condens. Matter, 12(40):8745–55, 2000.

[48] S.A. Studenikin, N. Golego, and M. Cocivera. Optical and electrical properties

of undoped ZnO films grown by spray pyrolysis of zinc nitrate solution. J. Appl.

Phys., 83(4):2104–11, 1998.

[49] T. Riedle. Herstellung von dunnen Cu-In Filmen und Charakterisierung mit dem

Rasterkraftmikroskop. Studienarbeit, TU Berlin, 1997.

[50] T.M. Friedlmeier, D. Braunger, D. Hariskos, M. Kaiser, H.N. Wanka, and H.W.

Schock. Nucleation and growth of the CdS buffer layer on Cu(In,Ga)Se2 thin films.

25th IEEE PVSC, pages 845–48, 1996.

[51] R.S. Mane and C.D. Lokhande. Chemical deposition method for metal chalco-

genide thin films. Mat. Chem. Phys., 65(1):1–31, 2000.

[52] A. Neisser. Gallium as an Isovalent Substitution in CuInS2 Absorber Layers for

Photovoltaic Applications. PhD thesis, FU-Berlin, 2001.

74 Bibliography

[53] M.C. Joliet, C. Antoniadis, R. Andrew, and L.D. Laude. Laser-induced synthesis

of thin CuInSe2 films. Appl. Phys. Lett., 46(3):266–67, 1985.

[54] N. Esser and J. Geurts. Raman spectroscopy. In G. Bauer and W. Richter, editors,

Optical Characterization of Epitaxial Semiconductor Layers, chapter 4. Springer,

1996.

[55] M. Cardona. Light Scattering in Solids, volume Volume I-VIII. Springer, 1975-

2000.

[56] Leon Brillouin. Diffusion de la lumiere et des rayons x par un corps transparent

homogene. Influence de l’agitation thermique. Annales de Physique, 17:88–122,

1922.

[57] C.V.Raman. A new class of spectra due to secondary radiation. Ind. J. Phys.,

2:387–399, 1928.

[58] M. Cardona and G. Guntherodt. Light Scattering in Solids II. Springer, 1982.

[59] I.H. Campbell and P.M. Fauchet. The effects of microcrystal size and shape on the

one phonon Raman spectra of crystalline semiconductors. Solid State Commun.,

58(10):739–741, 1986.

[60] K. Wakita, H. Hirooka, S. Yasuda, F. Fujita, and N. Yamamoto. Resonant Raman

scattering and luminescence in CuInS2 crystals. J. Appl. Phys., 83(1):443–446,

January 1998.

[61] J. Alvarez-Garcıa, A. Perez-Rodrıguez, A. Romano-Rodrıguez, T. Jawhari, J.R.

Morante, R. Scheer, and W. Calvet. Raman scattering structural evaluation of

CuInS2 thin films. Thin Solid Films, 387:216–218, 2001.

[62] G. Morell, R.S. Katiyar, S.Z. Weisz, T. Walter, H.W. Schock, and I. Balberg.

Crystalline phases at the p- to n-type transition in Cu-ternary semiconducting

films. Appl. Phys. Lett., 69(7):987–989, 1996.

[63] R. Hunger, M. Wilhelm, K. Diesner, J. Bruns, G. Lippold, K. Hinrichs, N. Esser,

H.J. Lewerenz, and R. Scheer. Phase Segregation in Epitaxial CuInS2 Films.

Proceedings of the 2nd PVSEC conference, 1998.

[64] K. Kampas, J. Spyridelis, and M. Balkanski. Far Infrared and Raman Optical

Study of α- and β-In2S3 Compounds. Phys. stat. sol. (b), 105:291–296, 1981.

[65] A.R. Verma and O.N. Srivastava. Crystallography for Solid State Physics. Wiley

Eastern Limited, 1981.

Bibliography 75

[66] K. Mathiak and P. Stingl. Gruppentheorie. Friedr. Vieweg und Sohn, 1968.

[67] J.W. Leech and D.J. Newman. How to use groups. Methuen and Co LTD and

Science Paperbacks, 1969.

[68] B.S. Wherrett. Group theory for atoms, molecules and solids. Prentice/Hall In-

ternational, 1990.

[69] D.L. Rousseau, R.P. Baumann, and S.P.S. Porto. Normal Mode Determination in

Crystals. Journal of Raman Spectroscopy, 10:253–290, 1981.

[70] J. Fritsch. Unpublished Results, 2001.

[71] J. Fritsch. Dynamische Prozesse und Phanomene an Halbleiteroberflachen. Ha-

bilitationsschrift, TU Regensburg, 2001.

[72] N.B. Colthup, S.E. Wiberley, and L.H. Daly. Introduction to Infrared and Raman

Spectroscopy. Academic Press, 1975.

[73] W. Kohn and L.J. Sham. Self-Consistent Equations Including Correlation and

Exchange Effects. Phys. Rev., 140(4A):A1133–38, 1965.

[74] Motohiko Ishii, Kenji Shibata, and Hiroshi Nozaki. Anion Distribution and Phase

Transitions in CuS1−xSex (x=0-1) Studied by Raman Spectroskopy. Journal of

Solid State Chemistry, 105:504–511, 1993.

[75] J. Klaer, J. Bruns, R. Henninger, K. Siemer, R. Klenk, K. Ellmer, and D. Braunig.

Efficient CuInS2 thin-film solar cells prepared by a sequential process. Semicond.

Sci. Technol., 13:1456–58, 1998.

[76] R. Scheer. Korrelation von Struktur und elektronischen Eigenschaften

mehrphasiger Cu-In-S Schichten fur die Photovoltaik. PhD thesis, TU-Berlin,

1993.

[77] R. Klenk, T. Walter, H.-W. Schock, and D. Cahen. Physical Vapor Deposition of

CuInX2 (X=S, Se) Thin Films: A Model for the Growth Mechanism. Solid State

Phenomena, 37-38:509–514, 1994.

[78] H. Richter, Z.P. Wang, and L. Ley. The one phonon Raman scattering in micro-

crystalline silicon. Solid State Commun., 39(5):625–629, 1981.

[79] G.L. Frey and R. Tenne. Raman and resonance Raman investigation of MoS2

nanoparticles. Phys. Rev. B, 60(4):2883–92, 1999.

[80] M. Gossla, H. Metzner, and H.-E. Mahnke. CuInS2 thin-films from co-evaporated

precursors. Thin Solid Films, 387:77–79, 2001.

76 Bibliography

[81] JCPDS. Powder Diffraction Files (PDF). International Centre for Diffraction

Data, 1998.

[82] T. Pagnier, M. Boulova, A. Galerie, A. Gaskov, and G. Lucazeau. Reactivity of

SnO2-CuO nanocrystalline materials with H2S: a coupled electrical and Raman

spectroscopic study. Sensors and Actuators B, 71:134–39, 2000.

[83] B.J. Wuensch. Determination, Relationships, and Classification of Sulfide Mineral

Structures. In P.H. Ribbe, editor, Sulfide Mineralogy, chapter 1. Mineralogical

Society of America, 1974.

[84] P. Derollez, R. Fouret, A. Laamyem, B. Hennion, and J. Gonzalez. Lattice dynam-

ics in copper indium diselenide by inelastic neutron scattering. J. Phys.: Condens.

Matter, 11:3987–3995, 1999.

[85] D.S. Su, W. Neumann, R. Hunger, P. Schubert-Bischoff, M. Giersig, H. J. Lew-

erenz, and R. Scheer. CuAu-type ordering in epitaxial CuInS2 films. Appl. Phys.

Lett, 73(6):785–787, 1998.

[86] D.J. Chakrabarti and D.E. Laughlin. Cu-s (copper-sulfur). In T.B. Massalski,

editor, Binary Alloy Phase Diagrams, chapter 2. ASM International, 1990.

Publications

Parts of this work have already been published:

Conference Proceedings

• Th. Riedle, Th. W. Matthes, A. Neisser, R. Klenk, C. Hinrichs, N. Esser, W.

Richter, M. CH. Lux-Steiner. Preparation of CuInS2 absorber layers by rapid

thermal sulfurization using H2S and DTBS. Proceedings of 16th EPVSEC, Glas-

gow, 713-716, 2000.

77

Curriculum Vitae

Thomas Riedle

Date and place of birth 01.02.1967, Plochingen

since January 1999 PhD student at the Technical University Berlin

and at the Hahn-Meitner-Institut Berlin

December 1998 university graduation: Diplom-Physiker

final thesis: Investigation of hydrogen in chalcopyrite

materials for solar cells

1991-1998 student of physics at the Technical University Berlin

1988-1990 civilian service with ”Action Reconciliation/

Services for Peace” in Tel Aviv / Israel

educational work for disadvantaged children

1986-1988 poly-technical school Stuttgart (Technische Oberschule),

high school graduation (Abitur)

1983-1986 vocational training in electronic-technician at

Gebr. Heller Werkzeugmaschinenfabrik

1973-1983 primary and secondary school, Wendlingen

78

Raman Spectroscopy for the Analysis of Thin CuInS Films€¦ · damentals of Raman spectroscopy...

Documents

Transcript of Raman Spectroscopy for the Analysis of Thin CuInS Films€¦ · damentals of Raman spectroscopy...